|
Day 2 Tuesday, February 10, 2026
|
| 08:00 AM - 09:40 AM |
Session 11A: Adaptive Traffic Control & Learning in ITS (Track: SRTS)
Edge-Enabled Detection for Adaptive Traffic Signal Control: A Comparative Study of YOLO-Based Models in Intelligent Transportation Networks
08:00 AM
Abrar S. Alhazmi, Abdelbaset Hamza and Abdallah Moubayed
ABSTRACT
Traffic congestion remains a major challenge in urban environments, requiring intelligent systems that can adapt to real-time conditions. Object detection is central to these systems, providing the vehicle and infrastructure data needed for effective traffic management. This paper compares five YOLO (You Only Look Once) models (v5, v8, v10, v11, v12) on real-world traffic video, evaluating inference time, detection accuracy, confidence scores, and memory usage. Results show clear trade-offs: YOLOv5 offers the fastest inference but with higher error rates, YOLOv8 maximizes sensitivity at the cost of more false positives, YOLOv10 delivers the strongest precision with fewest wrong guesses, and YOLOv11/12 provide stable predictions but slower speeds. These findings highlight that no single version dominates; instead, model selection should be guided by deployment needs, balancing speed, accuracy, and resource constraints in real-time traffic management.
Reinforcement Learning for Adaptive Traffic Signal Control: An Overview
08:20 AM
Aminu Yusuf, Tarek R. Sheltami and Ashraf Mahmoud
ABSTRACT
Traffic signal control is central to urban mobility because it directly influences congestion, travel time, and emissions. Traditional methods such as fixed-time and actuated control can handle normal conditions but often fail when traffic changes suddenly. Reinforcement learning (RL) has gained attention as a data-driven alternative that adapts policies by interacting with traffic environments. In recent years, both single-agent and multi-agent approaches have been explored, ranging from value-based to policy-based and actor-critic methods. This paper reviews how RL has been applied to adaptive traffic signal control, grouping the main approaches and highlighting their evaluation practices. The review shows that most studies still focus on efficiency measures such as delay and queue length, with safety and environmental factors less frequently addressed. Nearly all evaluations are done in simulation, with SUMO and CityFlow as the dominant platforms. Key challenges remain in handling multiple objectives, scaling to large networks, and bridging the gap between simulation and real-world deployment. By outlining current methods, their strengths and weaknesses, and the gaps that persist, this review points to the directions needed for RL to move from research to practice.
Adaptive multi-agent learning for infrastructure-aware ITS: the IMER data-processing approach
08:40 AM
Mayssa Hamdani, Nafaa Jabeur, Ansar Yasar, Fatma Outay and Li Li
ABSTRACT
The performance of Intelligent Transportation Systems (ITS) critically depends on accurate and efficient road-condition monitoring. This paper presents IMER (Inspect–Map–Eliminate–Reduce), a novel AI-driven data-processing framework that extends the traditional Map-Reduce paradigm for infrastructure maintenance. IMER integrates confidence-based validation, redundancy elimination, and severity prioritization to enhance data quality and decision efficiency. Implemented within a multi-agent architecture, IMER enables autonomous agents to inspect, classify, and fuse multi-source road data in real time, supporting predictive and adaptive maintenance planning. Simulation results using augmented pothole datasets demonstrate a 39.9 % reduction in redundant reports and 39.8 % fewer false positives. These findings highlight IMER’s potential to advance data-driven, resilient, and sustainable road-infrastructure management for next-generation ITS.
Graph-Based Causal Machine Learning for Self-Adaptive Control under Context Shifts in Smart Urban Mobility
09:00 AM
Fernando Antonio Montoya Cáceres, Hernán Astudillo, Mauricio Hidalgo, Sergio Montes-León and César Viloria-Núñez
ABSTRACT
Urban mobility services based on pickup–delivery routes operate under changing conditions: traffic, operational corridor constraints, schedules, and demand fluctuations continuously alter the context in which control decisions are executed. In many deployments, these decisions rely on fixed policies and prediction models trained on a “typical” context, which lose accuracy and effectiveness when the situation deviates from that scenario. This paper addresses how to design control mechanisms that keep the temporal performance of the service robust to context shifts, allowing the management software to adapt at runtime without manual tuning or purely correlational models. We propose a self-adaptive control approach guided by graph-based causal machine learning. The system is represented through causal graphs specific to each operational unit, instantiated here at the route level for a public school transport service, linking context variables, control decisions, and temporal outcomes. On these graphs, we train CACM-based predictive models that distinguish causal from spurious attributes, and integrate the resulting model into a MAPE-K loop that selects, according to the estimated context, from a bounded set of timing adaptations implemented via context-oriented programming (COP). The approach is validated in a simulation environment adjusted with historical records from the school transport platform, with 25,600 trips distributed across five contexts that combine demand level and route configuration. Results show that CACM maintains predictive performance comparable to a standard ERM baseline while reducing delay variability and headway irregularity when used to drive adaptive policies. The contribution is a practical framework that combines graph-based causal machine learning and COP-based self-adaptation for urban transport services, together with an initial assessment of its benefits over non-adaptive and correlational alternatives.
Delay-Aware Car-Following Model for Mixed Traffic: Impacts of Automated Vehicles on Traffic Dynamics
09:20 AM
Daud Khan, Waheed Imran, Kamil Wittek, Katarzyna Markowska and Susilawati Susilawati
ABSTRACT
Recent studies on the yet-to-be-realized integration of autonomous vehicles (AVs) into conventional traffic, i.e., with human-driven vehicles (HVs), identify heterogeneous temporal responses as a distinguishing characteristic, which can alter traffic stability, safety, and disturbance propagation. The classical car-following model, Intelligent Driver Model (IDM), characterizes traffic based on instantaneous reaction and fixed headways, and perception response delays of drivers. To address this limitation, this study proposes a temporal extension of the IDM that embeds a source term, i.e., effective headway, to model the desired dynamic gap. The proposed extension has been tested on a 1000-m single-lane hypothetical corridor under varying levels of automation (0%, 60%, 100%). Further, a controlled deceleration perturbation was applied to evaluate disturbance propagation, recovery dynamics, and string stability. Results show that HVs traffic exhibits pronounced speed drops, amplified oscillations, and slow return to equilibrium due to long reaction delays. The results demonstrate that incorporating realistic reaction delays and buffer-based spacing yields a more accurate representation of mixed traffic and highlights the stabilizing role of AVs as temporal headways decrease. The proposed framework provides an extensible basis for multi-lane applications, control-policy design, and empirical calibration using AV-HV trajectory datasets.
|
| 08:00 AM - 09:40 AM |
Session 11B: Mathematical Optimization and Algorithmic Decision Models in Logistics (AOIL)
Learning-Assisted Optimization for the Stochastic Service Network Design Problem
ABSTRACT
Service Network Design (SND) problems arise when consolidation carriers plan vehicle or vessel capacities for their transportation networks for an upcoming operating season. The two-stage Stochastic SND (SSND) further explicitly considers uncertain demand: carriers commit to offering transportation services in advance (first stage) and may later adapt by procuring additional costlier spot-market capacity (second stage). While more realistic, SSNDs are computationally challenging. In this work, we study the use of machine learning, in particular artificial neural networks, to assist the solution of two-stage Stochastic Service Network Design (SSND) problems. The aim is to learn a predictive model that estimates the second-stage performance of candidate first-stage SND solutions under uncertain demand. This learned “simulator” can then be embedded into optimization/heuristic frameworks, forming hybrid approaches that potentially increase the tractability of instances. Our results show that that effective learning can be achieved with an estimation gap (between the actual expected cost of a solution and its estimated value) of 6% on average, and that the quality of estimation decreases with the instance size (for a fixed training set size).
Efficient Path Planning in Obstacle-Dense Grids via Advanced Optimization
08:20 AM
Hamza Tahiri, Mohamed Amine Tahiri, Hicham Karmouni, Mhamed Sayyouri and Mohamed Abouhawwash
ABSTRACT
Path planning on cluttered grids challenges deterministic methods due to scalability and trajectory irregularities. We adapt three recent metaheuristics Moss Growth Optimization (MGO), Rime Optimization (RIME), and Polar Lights Optimizer (PLO) to discrete waypoint-based navigation with obstacle penalties and path-shape regularization. On a 20×20 map with 52 obstacles, we evaluate cost, length, turns, convergence, and runtime over multiple trials. MGO delivers the most stable performance with smooth, low-turn trajectories and the lowest average cost; PLO attains the best single cost through strong late-stage exploitation, albeit with slightly higher runtime and irregularity, RIME converges fastest initially but stalls early, yielding higher variance. Statistical tests confirm the trade-offs among stability, optimality, and speed. We conclude with ablations and guidelines on algorithm selection and outline extensions to larger, dynamic maps and kinodynamic robots
A Carbon Aware Two-Echelon Location-Capacity-Routing Problem with Outsourced and In-House Delivery
08:40 AM
Hudaifah, Andriansyah Hamid and Mohammad Abdel-Aal
ABSTRACT
Green supply chain management has become increasingly important to reduce the environmental impacts of logistics activities. The expansion of business operations has led to the widespread adoption of two-echelon supply chain structures across various sectors. This study proposes a multi objectives mathematical model for the Green Two-Echelon Location Routing Problem, incorporating depot sizing decisions as part of the optimization process (2E-LCRP). The proposed model considers two objectives: (1) total cost, including cost of depot establishment, transportation, and vehicle usage across both echelons, and (2) total carbon emission, accounting for both transportation and depot operational emissions. Lexicographic optimization is employed to effectively balance these objectives. The model accommodates both outsourced and in-house delivery operations. Experimental results indicate that the proposed approach successfully manages the trade-off between cost and environmental impact. This work advances the field of green scheduling by providing a comprehensive framework for carbon emission assessment and extending the two-echelon location routing problem to integrate depot sizing decisions.
Making informed inventory decisions, a revisit of the continuous review model, where back orders are allowed
ABSTRACT
Inventory management is the cornerstone of modern supply chain systems. It works as an intermediate hub that facilitates the smooth transition of products between several supply nodes. In inventory models, the total cost is random, and it rarely matches the expected cost, leaving management unaware of actual cost fluctuations monitored and reported by accounting systems. The mean and variance approach has been employed to address this issue. However, most existing work either considers the decision maker's attitude within the problem or treats it as a multi-objective problem, developing an efficient solution frontier. Unlike earlier studies, this research considers the mean and variance as a single-objective function, allowing the evaluation of the total cost at various risk levels (α). This approach allowed determining the probability distribution of the total cost, which led to a model that defines its boundary limits. Knowing these limits should help management make better-informed inventory decisions. The findings also show that the proposed and expected cost models follow the same optimal policy under certain conditions. These results are confirmed using a numerical example and validated by a simulation.
Sustainable Supplier Selection under Quantity Discounts: An Epsilon-Constraint Model for Cost and Water Footprint Reduction
09:20 AM
Osamah Mohammed, Sani Ahmad and Mohammad Abdel-Aal
ABSTRACT
This study proposes a multi-objective mixed-integer nonlinear optimization model for sustainable supplier selection under quantity discounts, aiming to minimize total cost and water footprint across a multi-echelon supply chain. The mixed-integer nonlinear model integrates purchase, process, and transportation costs with blue, green, and grey water footprints to capture economic and environmental trade-offs. To address nonlinearity and computational complexity, the Epsilon-constraint method is applied to transform the multi-objective problem into single-objective subproblems. Numerical experiments demonstrate the model’s effectiveness in identifying Pareto-optimal solutions that balance cost efficiency and sustainability. The results highlight how slight relaxations in water footprint constraints can significantly reduce total cost, providing insights into the cost–environmental trade-off in supplier selection.
|
| 08:00 AM - 09:40 AM |
Session 11C: Intelligent & Autonomous Transport Systems in Cognitive Cities (Track: SRTS)
AI-Powered Monocular Perception for Collision Avoidance and Resilient Navigation in Unmanned Surface Vessels
08:00 AM
Faisal Alrebeish, Hesham Shageer and Mohammed Alsayed
ABSTRACT
This paper introduces a robust, AI-driven monocular perception framework for Unmanned Surface Vessels (USVs). We propose a deep learning–based object detection pipeline optimized for real-time vessel recognition and collision avoidance using lightweight, single-camera vision systems. The primary contribution of this work is the first systematic evaluation of YOLOv12’s robustness in monocular USV perception under simulated maritime challenges such as glare, motion blur, and varying illumination conditions. Our framework benchmarks four state-of-the-art object detectors—YOLOv8, YOLOv11, YOLOv12, and RF-DETR—using an augmented Sea Vessels Dataset to assess resilience and performance. Experimental results show that YOLOv12 achieves the best balance between precision (93.8%) and recall (78.8%), enabling safe and reliable operation of small autonomous vessels. These findings demonstrate that, with advanced AI models, monocular vision can deliver trustworthy situational awareness and decision support for next-generation connected and resilient maritime mobility systems.
Intelligent Fault and Cyber Intrusion Detection Framework for Cyber Physical Systems: Integration of Electric Vehicle with Smart Grid
08:20 AM
Zahid Riaz, Salman Habib, Bilal Khan and Muhammad Majid Gulzar
ABSTRACT
Electric vehicles (EVs) play a critical part in the transition to environmentally friendly transportation while minimizing greenhouse gas emissions. However the increasing integration of Electric Vehicles (EVs) into modern power grids introduces the reliability, security, and resilience challenges in transportation energy networks. The bidirectional power flow between EVs and the grid (V2G) produce a complex cyber intrusionphysical network susceptible to electrical faults and cyber threats. this research proposed AI can perform early fault detection and cyber attack identification in integrating EV to smart grid (EV-Smart Grid) systems, ensuring secure and reliable energy transfer. This paper presents an intelligent framework for fault detection and cybersecurity enhancement in EV–Smart Grid systems. This paper proposes an AI-based hybrid framework combining Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM) networks, and Transformer attention mechanisms for real-time fault and cyber-anomaly detection. The system processes high-resolution voltage, current, and frequency measurements collected from EV charging systems. The model is trained and validated on simulated representing typical fault events and cyber-injected anomalies in grid connected EV structure. The results demonstrate high accuracy and robustness in detecting faults and cyber anomalies in electrical signals. This work contributes toward building secure, intelligent, and resilient smart mobility structures by merging AI-based fault diagnostics with cybersecurity monitoring, supporting the realization of trustworthy electric mobility and connected smart grid systems.
Virtual Testing of Autonomous Driving System with Cooperative Perception V2X Communications For Enhancing Safety of Pedestrians in Urban NLOS Scenarios
08:40 AM
Mohammed Shabbir Ali, Pierre Merdrignac and Mohamed-Cherif Rahal
ABSTRACT
Urban autonomous driving systems face significant safety challenges in non-line-of-sight (NLOS) situations, particularly when pedestrians suddenly emerge from behind occlusions. Cooperative perception, enabled by ETSI-standard Cooperative Perception Messages (CPMs), addresses this limitation by extending situational awareness beyond onboard sensors. This paper presents a virtual testing study of an autonomous driving system (ADS) equipped with CPM-based cooperative perception for a darting-out pedestrian scenario. A simulation environment integrating OMNeT++, Veins, and SUMO is developed, featuring a camera-based perception model with occlusion handling and an ADS control module that fuses CPM data for speed adaptation and emergency braking. This virtual-first approach enables rapid scenario exploration and scalable evaluation of ADS and V2X performance. Key Performance Indicators (KPIs) assess perception coverage, CPM delivery ratio, communication latency, mobility adaptation, and safety outcomes. Results show that CPM-augmented perception identifies the pedestrian up to two seconds earlier than onboard sensing alone. Connectivity evaluation indicates high CPM delivery ratios with low latency in ideal conditions, supporting reliable cooperative perception. Time-to-Collision and braking profiles confirm smooth and safe ADS responses, while edge cases highlight operational limits where collisions remain unavoidable, providing valuable scenarios for subsequent high-fidelity testing. Overall, the study demonstrates that virtual-first testing is an effective and scalable method for developing and validating CPM-based cooperative perception in urban autonomous driving.
Balancing Berth Utilization to Reduce Anchorage Waiting Time for Port Terminals in Coastal Cities
ABSTRACT
Effective seaport management is vital for sustaining international trade and driving the economic growth of coastal cities. Seaport capacity is usually defined by the size and number of berths. Therefore, the level of effective utilization of sea berths will greatly affect a seaport’s management and consequently the average waiting time at anchorage, which will result in demurrage costs. This paper defines six critical factors that should be maintained to have effective utilizations of sea berths. The case of eight berths is studied to minimize the average waiting time at anchorage for a petrochemical (liquid) Terminal. The focus of this paper is to analyze data of product throughput, number of visits per parcel and berth characteristics. The results suggest that products should be reallocated in berths and additional outlets should be added to prevent vessels (parcels) from waiting in queue for a berth and reduce the shifting of vessels (moving from one berth to another), which increases the idle time. The forecasting plan, taking into consideration factors that influenced the port’s overall efficiency, improved the overall occupancy of berths. Improvement was realized on the level of overall occupancy of berths and the average waiting time at anchorage which will improve the demurrage time.
GIS-based Multi-Criteria Decision Analysis for Optimal Parcel Locker Site Selection in Eastern Province, Saudi Arabia.
09:20 AM
Md. Aqib Aman, Ahmad Al-Hanbali, Ahmed M. Attia and Baqer Al-Ramadan
ABSTRACT
E-commerce growth has increased the need for effective last-mile delivery systems. It makes parcel-locker placement an important planning task. This work presents a GIS-based Multi-Criteria Decision Analysis (MCDA) method to identify suitable locker sites in Saudi Arabia’s Eastern Province. ArcGIS Pro tools such as Euclidean distance, reclassification, weighted overlay, and raster calculations were applied to combine spatial and socio-economic factors. Key criteria included accessibility, proximity to residential zones, transport networks, and commercial areas. The resulting suitability map highlights optimal locations to improve delivery efficiency and service coverage for urban logistics planners.
|
| 08:00 AM - 09:40 AM |
Session 11D: Smart Mobility, Driver Monitoring, and Road Safety Analytics (Track: CHTML)
Modeling Driver-Caused Road Crash Severity Using SHAP-Enhanced Machine Learning
08:00 AM
Makides Damene, Muhammad Abdullah, Hassan Al-Ahmadi and Muhammad Aamir Basheer
ABSTRACT
Road traffic crashes continue to be a significant global issue, causing millions of fatalities and considerable social and economic losses annually. Saudi Arabia consistently reports one of the highest incidences of traffic-related crashes globally. Although numerous prior studies have approached crash prediction mainly as a classification issue centered on the probability of occurrence, few have investigated the fundamental relationships among the factors affecting crash severity, especially those related to driver behavior. This study examined driver-cause road traffic crashes throughout the Kingdom of Saudi Arabia (KSA) utilizing crash data from 2017 to 2022 sourced from the Ministry of Transport (MOT). Injury severity was categorized into three classifications: Fatal, injury, and property damage only. A variety of machine learning models were developed, including Decision Tree, Random Forest, AdaBoost, Gradient Boosting, and XGBoost. Among these, the XGBoost model performance well. Shapley Additive Explanations (SHAP) were examined to ascertain the global contributions of features for improved interpretability. Global SHAP analyses revealed that Accident Type, Accident Cause, Road Surface Condition, Road Type, and Vehicle Type were the most important predictors across all severity levels. Dependence analysis further uncovered strong feature interactions: speeding and freeway road type had the greatest impact on fatal crashes. Fatigue- and sleep-related causes, especially under nighttime conditions, were the contributors to injury crashes, while good road surface conditions and freeway road type were associated with property-damage-only outcomes. The integrated predictive and interpretable modeling framework provides, data-driven insights to inform targeted interventions aimed at reducing road accident severity across Saudi Arabia.
A Review of Sociodemographic Influences on Car-Following Behavior
08:20 AM
Zahid Hussain, Muhammad Baqir and Tufail Ahmed
ABSTRACT
Car-following behavior significantly influences traffic flow dynamics and represents a major factor in rear-end collision occurrences during unexpected events. Understanding the impact of human factors, which are widely recognized as the predominant cause of road crashes, is essential for improving traffic safety. Therefore, this review study synthesized the literature on the effects of heterogeneity in human characteristics on drivers' behavior during car-following situations. The literature search was conducted in the Scopus database, and 21 studies were selected for review. This study summarizes the results related to the influence of age, gender, driving experience, and cultural differences on car-following behavior. The results revealed that younger drivers showed faster reaction times but faced higher rear-end collision risk due to speeding and tailgating behavior. Regarding gender differences, the findings revealed mixed patterns. While male drivers generally exhibited more aggressive car-following characteristics with shorter following distances and higher speeds. The relationship between gender and crash risk showed inconsistencies, as some studies reported higher collision risk for females despite more conservative behavior. Driving experience appears to influence car-following behavior, with experienced drivers potentially exhibiting overconfidence and aggressive driving tendencies. Additionally, the limited research on cultural differences highlights the need to examine car-following behavior among culturally diverse driving populations. This would provide valuable insights and help calibrate car-following models to be more universally applicable. Moreover, the review underscores the need for interventions to address unsafe car-following behaviors and enhance traffic safety.
Analysis of driver expectancy in cut-in maneuvers: A Virtual reality driving simulator study
ABSTRACT
Autonomous vehicles (AVs) will, for the foreseeable future, share roadways with human-driven vehicles, making safe and predictable lane-change interactions a critical human-factors and safety challenge. However, how human drivers cognitively interpret and respond to abrupt AV cut-in maneuvers remains poorly quantified, with most prior research focusing on control and operations rather than driver expectancy. To the best of our knowledge, this study presents the first systematic analysis of driver expectancy violations in AV cut-in scenarios using an immersive Virtual Reality Driving Simulator (VRDS). It addresses a fundamental gap by linking psychophysical measures of driver surprise to specific longitudinal interaction gaps, cut-in directions, and traffic contexts, thereby providing novel evidence for human-centred AV behavior design. Fifty-one licensed drivers in Saudi Arabia completed a VRDS experiment in which an AV executed single-lane cut-ins on a three-lane freeway under controlled conditions. Cut-ins originated from the left or right adjacent lane, with relative speeds of ±5 mph, with or without a leading vehicle in the target lane, and with temporal gaps of 0.5, 0.75, and 1.0 s between the subject vehicle and the cut-in vehicle. Driver expectancy violations were recorded on a five-point scale and analyzed using a Bayesian multilevel ordinal regression model. Gap and cut-in direction (combined with relative speed) emerged as the most influential predictors. Across scenarios, a 1.0 s gap rarely violated driver expectancy, with at least 85% of participants reporting no or only slight surprise, whereas 0.5 s gaps predominantly produced strong or complete expectancy violations. Left-lane cut-ins at higher speed than the subject vehicle were generally anticipated, while right-lane cut-ins at lower speed generated substantially greater violations, slightly intensified by the presence of a leading vehicle. These findings yield actionable parameters for AV lane-change algorithms, mixed-traffic safety assessment, and the development of replicable VRDS-based protocols for evaluating future AV–human interaction designs.
From Posture to Performance: Continuous Cognitive Load Estimation of Drivers Using Multi-Modal Anatomical Keypoints
09:00 AM
Golam Sakaline, Syed Md Sadman Rafid, Moshiur Ahmed and Ahmet Kolus
ABSTRACT
Driver distraction and cognitive overload are significant contributors to road accidents; however, most vision-based driver monitoring systems focus on classifying observable behaviors rather than estimating underlying cognitive workload. This study presents a non-contact, interpretable framework for continuous cognitive load (CL) estimation that relies solely on 2D pose, face, and hand keypoints extracted from realistic driving images. Using the Multi-Class Driver Behavior Image Dataset collected in Ashulia, Dhaka (2024), five driver behavior categories are mapped to normalized CL scores (ranging from 0 to 1), reframing the problem as a regression task to capture the continuous spectrum of cognitive effort. MediaPipe is utilized to extract body posture, head orientation, and hand-movement keypoints, which are then normalized and combined into a unified feature vector. The proposed Multi-task Learning (MTL) architecture, built upon a shared Multi-layer perceptron (MLP) backbone, simultaneously predicts discrete driver behavior and continuous CL using two task-specific output heads. The results show competitive predictive performances with Pearson of 0.58, MSE of 0.052, and R² = 0.34, too, meaning that predicted and ground-truth CL scores are very close. Analysis of the importance of features shows that the biggest predictors of cognitive effort are cues based on body position such as torso lean, head tilt, and hand placement. It generates continuous load distributions despite being trained with discrete behavioral labels while maintaining within-class variability. This work provides the first example of a privacy-preserving, low-cost and explainable alternative to intrusive physiological sensing by combining behavioral action recognition with continuous cognitive state estimation. To enable immediate integration into Advanced Driver Assistance Systems (ADAS), driver monitoring modules, and pipelines for detecting the fatigued or distracted driver, the method is compact and operates in real-time.
Driver Yawning Monitoring for Connected Mobility Using Spatiotemporal Hybrid Transformer
09:20 AM
M Faisal Nurnoby and El-Sayed M. El-Alfy
ABSTRACT
Driver fatigue is a major cause of road accidents. Yawning is an early and reliable cue of drowsiness. This paper presents a vision-based system for driver yawning monitoring, which can be an integral part in connected vehicle technology to share real-time fatigue clues with surrounding vehicles. Our method uses a Spatiotemporal Hybrid Transformer (HYVTR) that models local facial appearance and long-range temporal cues to detect driver yawning. It reduces false positives from talking and other mouth actions including laughing. We also propose a simple labeling protocol based on mouth-aspect-ratio (MAR) frequency thresholding. This protocol processed long video files into valid and consistent class label for balanced training data generation. We evaluate our suggested approach on the YawDD and NTHU-DDD (Yawn) datasets and it achieves 93.40% accuracy on YawDD and 99.34% accuracy on NTHU-DDD. We compare against facial landmarks based spatiotemporal Swin-T and ST-GCN baselines. Our model outperforms all baselines on both datasets. The results suggest strong robustness under pose and illumination changes. The system supports practical deployment for smart and safe transportation.
|
| 08:40 AM - 11:40 AM |
Session 12: QCAR Competition |
| 08:40 AM - 10:20 AM |
Session 13: Poster Session
Policy Alignment for Electric Mobility in Saudi Arabia: An L₁ Norm Decomposition Approach
Amjad Ali, Md Tasbirul Islam, Sikandar Abdul Qadir and Muhammad Shahid
ABSTRACT
Saudi Arabia has set ambitious electric mobility targets under Vision 2030, yet the extent of policy alignment with these objectives remains uncertain. This study applies an L₁ Norm (Manhattan Distance) based decomposition framework to quantify policy misalignment across five pillars: adoption targets, charging infrastructure, consumer incentives, domestic manufacturing, and regulations and standards. Results indicate that Saudi Arabia’s EV policy framework exhibits a normalized deviation of 0.68, implying only 32% cumulative alignment with the Vision 2030 benchmark. Decomposition analysis reveals that charging infrastructure, consumer incentives, and regulatory standards each account for 23.5% of the total misalignment, while adoption targets contribute 17.6%, and domestic manufacturing shows relatively lower shortfalls at 11.8%. These findings highlight that demand-enabling policies remain the most critical barriers to achieving EV mobility goals. By pinpointing the relative contribution of each pillar to the overall policy gap, the study provides a diagnostic basis for prioritizing reforms and sequencing interventions. The L₁ decomposition framework thus offers a transparent and replicable tool for evidence-based policymaking in the Saudi context and can be extended to other emerging EV markets facing similar implementation challenges.
A Smart Digital Transformation for Enhancing Operational Efficiency in Manufacturing
Abdulkhaliq Al-Shamrani, Awsan Mohammed, Ahmed Ghaithan and Ahmed Attia
ABSTRACT
Digital Transformation (DT) has become a strategic priority for manufacturing firms seeking to enhance competitiveness, operational efficiency, and sustainability. This paper investigates the implementation of digital transformation in a steel manufacturing plant, exploring the integration of artificial intelligence (AI), machine learning (ML), the Internet of Things (IoT), and big data analytics. The study uses a structured approach that includes assessment, diagnosis, model development, and implementation phases. Findings reveal that DT enhances productivity, reduces costs, and improves equipment efficiency, though challenges such as cost, skill gaps, and cybersecurity persist. The results contribute to the understanding of how digital transformation strategies can be operationalized in traditional manufacturing contexts.
Health-Critical Event Detection in Hajj Crowds Using Action Recognition and Facial Affect Analysis
Salma Aldossary, Sadam Al-Azani and Wadha Almattar
ABSTRACT
Crowd anomaly detection is a critical capability for public safety in large-scale gatherings. Although a wide range of anomaly-detection methods has been proposed, most existing work concentrates on motion irregularities and crowd-density estimation. In contrast, comparatively little attention has been given to detecting individual-level, health-related indicators. This motivates us in the paper to present a computer-vision-based monitoring framework for detecting health-critical behaviors and early distress cues in large-scale pilgrimage crowds, using object detection and facial affect analysis. As existing datasets lack a specific focus on health-related indicators, we defined nine health-focused classes and developed Hajj-Health dataset. This dataset extends HAJJv2 by incorporating relevant objects from it, alongside newly collected and AI-generated images representing two additional classes: Raise Hand for Help and Receiving Medical Assistance, to capture health-related behaviors during Hajj. Different YOLO versions, including YOLOv8, YOLOv9, YOLOv10, and YOLOv11, were trained and evaluated for action recognition. The experimental results demonstrate that YOLOv8 and YOLOv9 achieve the overall highest performance in terms of precision, recall, and $mAP$. In addition, the framework detects faces and identifies stress-related affect using RetinaFace for face detection and DeepFace for emotion recognition. These findings indicate that the proposed approach can reliably detect standard crowd behaviors, subtle health indicators, and stress-related cues, supporting early intervention in dense gatherings. This proposed framework lays the foundation for real-time, health-aware crowd monitoring during Hajj and similar mass events within cognitive city environments.
Robust Terminal Sliding Mode Control of Robotic Manipulators in Logistics Automation
Nidhal Khorchani, Rabeb Benjemaa and Nezar M. Alyazidi
ABSTRACT
Robotic manipulators are a core enabler of smart warehousing and resilient logistics, yet they operate under modeling uncertainty, payload variability, and external disturbances that degrade tracking accuracy and energy efficiency. We develop an adaptive non singular terminal sliding mode controller (ANTSMC) that achieves finite-time trajectory tracking under bounded disturbances while explicitly reducing chattering and control energy to throughput and actuator longevity in green supply chains. The design introduces (i) a terminal sliding surface that yields an explicit settling-time bound and avoids singularities, and (ii) an adaptive switching-gain law with a dead-zone correction that scales with error magnitude, improving robustness without excessive control effort. Lyapunov analysis establishes finite-time convergence and a bound on the integrated torque. Sample simulations showed that the proposed RTSMC compared favorably in terms of tracking errors, robustness, and disturbance rejection against the sliding mode controller and PID controller. Overall, the findings highlight the potential of this RTSMC to increase throughput, reliability, and efficiency of operations in advanced logistics.
A Comparative Study of Two-Phase Clustering–TSP and DQN-Enhanced Metaheuristic Approaches for CVRP
Rahaf Alzahrani, Alaa Khamis and Zead Saleh
ABSTRACT
This paper presents an integrated optimization framework for solving the Capacitated Vehicle Routing Problem (CVRP) in real-world beverage distribution logistics. The proposed approach combines clustering techniques for spatial decomposition of customer locations with adaptive metaheuristics, including Adaptive Genetic Algorithms (AGA), Adaptive Simulated Annealing (ASA), and reinforcement learning through Deep Q-Networks (DQN). Clustering reduces computational complexity by grouping geographically proximate delivery points, while the adaptive algorithms dynamically adjust key parameters based on solution feedback to enhance search performance. Applied to a real-world distribution network, the framework achieved notable improvements in routing efficiency, vehicle utilization, and total travel distance. Compared to conventional heuristics, the integrated method demonstrated superior scalability, adaptability, and responsiveness to dynamic constraints, making it well-suited for large-scale logistics operations. The results underscore the potential of combining artificial intelligence and metaheuristic techniques to enable intelligent routing decisions, optimize fleet management, and reduce operational costs in practical distribution scenarios.
PWM-Based Intelligent Speed Control of BLDC Motors for Sustainable Electric-Vehicle Drives
Ziad Shoeib, Khaled Bin Gaufan and Nezar M. Alyazidi
ABSTRACT
Due to their high efficiency and compact design as well as their favorable torque-to-power ratio, permanent magnet brushless (BLDC) motors are universally acknowledged to be critical enablers in future generations of motion and automation systems. They serve as primary actuators in electric vehicles, e-scooters, and logistics robots, where precise and robust speed control remains challenging because of nonlinear dynamics and parameter variations. This paper analyzes three strategies of speed control of BLDC motors: classical PID control, fractional-order PID control (FOPID), and adaptive neuro-fuzzy inference systems (ANFIS). The proposed framework, implemented in MATLAB/Simulink, evaluates robustness, energy efficiency, and adaptability under dynamic conditions. Simulation results show that while the PID controller ensures acceptable transient performance, the FOPID achieves superior steady-state accuracy and disturbance rejection. The ANFIS controller provides adaptive learning capability suited for AI-driven mobility systems. Overall, the study demonstrates that the FOPID controller offers a reliable and high-precision solution for BLDC motor control in intelligent transportation and automated logistics applications.
GlassSkyDeliver: Design and Evaluation of a Smart, Sustainable, and Automated Urban Transport System for Last-Mile Delivery.
ABSTRACT
Urban logistics has experienced a rapid transformation driven by the exponential growth of e-commerce and on-demand delivery services. However, the reliance on conventional ground-based systems, mainly cars and motorcycles, has intensified urban congestion, greenhouse gas emissions, and safety risks, particularly in densely populated cities. These challenges underscore the need for a more sustainable and technologically advanced delivery framework aligned with global environmental objectives and national visions such as Saudi Vision 2030. This study introduces and evaluates the GlassSkyDeliver system, a pioneering overhead delivery network composed of transparent elevated tubes through which autonomous pods transport goods between vendors and customers. The system’s architecture integrates three main components: the vertical transfer Line for vertical movement and secure access, the horizontal track line as the primary elevated transport corridor, and intelligent delivery pods that operate autonomously with real-time routing, Global Positioning System (GPS) tracking, and digital security verification. The design aims to achieve zero-emission operations, eliminate ground-level congestion, and enhance urban aesthetics by leveraging existing infrastructure such as streetlight poles. Through its automated workflow and renewable power sources, GlassSkyDeliver provides a sustainable alternative to conventional delivery systems while ensuring speed, safety, and reliability. A comparative analysis between GlassSkyDeliver and traditional modes such as cars and motorcycles demonstrates the superiority of the proposed system in reducing environmental impact, improving energy efficiency, and enhancing overall delivery performance. The study highlights the feasibility of implementing elevated, automated logistics networks as a core component of future smart cities.
Dilated Convolution and Adaptive Upsampling-Based FCN Framework for High-Precision Semantic Segmentation in Self-Driving Vehicles
Atta Rahman, Bibi Saqia and Muhammad Zubair
ABSTRACT
Autonomous driving needs highly precise and real-time vision systems to ensure safe navigation in complex, dynamic, and changing driving environments. The main flaw of this area lies in properly identifying and classifying nearby objects, such as lanes, pedestrians, roads, and vehicles. This retaining spatial reliability across a promptly changing condition. The limitations of existing semantic segmentation approaches are insufficient object recognition, containing road boundary and lane; low accuracy in congested or low-light situations; large processing time; and harm of spatial data as a consequence of feature flattening. To report these limitations, we suggest a progressive Fully Convolutional Network (FCN)-based semantic segmentation framework for self-governing driving assignment. The proposed model presents important contributions to overcome the inadequacies of earlier work. We employed a thorough convolution to expand the open domain without increasing the number of parameters of the structure. The model perceives both fine-grained local features and global data efficiently. Then, we utilized an adaptive up-sampling method to enhance segmentation structures. This allows precise boundary identification between prudently spreading substances even under multi-layered biological conditions. The proposed model offers excellent spatial detail and computational capability for semantic segmentation with these advancements. Experiments were conducted on the Lyft Udacity challenge dataset to verify the performance of the proposed model. The model achieved 93.04% accuracy, outperforming state-of-the-art models. These results demonstrate that the suggested FCN-based model significantly improves scene understanding for self-driving vehicles, supporting more consistent navigation in real-world conditions.
Evaluating Neighborhood Park Quality: A Synthesis of Theory and Practice through Case Studies
Sharif Tousif Hossain, Mohammad Sulaiman Abu-Zaghlan and Ahmad Abdulqader
ABSTRACT
The paper will examine and integrate the key qualities of a good neighborhood park by conducting a thorough literature review and also use them in a practical evaluation of the Thuqbha Park located in Al Khobah, Saudi Arabia. The research, based on more than twenty peer-reviewed studies, indicates that three important quality dimensions are identified, namely, accessibility, safety and security, and psychological comfort. The following attributes were operationalized by the use of on-site observational tests and a questionnaire survey. The analysis shows that Thuqbha Park has a positive attribute of social activity and simple facilities, but it does not have a variety of vegetation, inclusive design, and environmental comfort. The paper ends with practical suggestions on how to advance the quality of neighborhood parks in swiftly developed areas with a focus on the user-oriented design, the inclusive planning, and the ongoing monitoring. The study will help the researcher with a scalable model of evaluating and improving the open spaces of the population, specifically in the Saudi environment.
Autonomous Indoor Rescue Drone Guided by Sound and LLMs
Rawan Bakro, Raghad Al-Johani, Asma Al-Arfaj and Hussein Samma
ABSTRACT
In high-risk indoor emergencies such as fires, autonomous drones can play a vital role in accelerating and improving rescue operations. This research introduces a sound-guided indoor rescue drone enhanced by Large Language Models (LLMs) to enable intelligent perception and decision-making. The system development is collaboratively undertaken by two students: Rawan focuses on sound recognition and LLM-based navigation reasoning, while Raghd leads vision processing and drone navigation. By fusing auditory and visual data, the drone achieves robust situational awareness and autonomous path planning in dynamic environments. Significant milestones include the successful integration of sound detection, scene analysis, and LLM-driven navigation, marking a step toward more adaptive and context-aware rescue robotics.
Design of Finite Time Sliding Mode Control for Stable DC Bus and Current Tracking in a Multisource Hybrid Electric Vehicle
Usama Faiz, Waqar Uddin, Muhammad Khalid and Kamran Zeb
ABSTRACT
The increasing demand for clean and efficient transportation has shifted research focus towards hybrid electric vehicles. This study focuses on designing a nonlinear controller for a multisource hybrid electric vehicle comprising a fuel cell, battery, supercapacitor, and photovoltaic panels. DC converters act as an interface between sources and the DC bus. An FTSMC controller is designed and developed to ensure accurate current tracking for all sources and effective regulation of the DC bus voltage. The stability of the proposed controller is verified using Lyapunov stability analysis, and its performance is assessed using MATLAB/Simulink. The EUDC cycle has been used as a load profile for the proposed HEV. The results demonstrate fast and robust control performance under varying operating conditions, confirming the controller’s effectiveness for multisource HEV applications.
Boosting Classification of Electric Vehicles from Charging Patterns for Smart Mobility and Sustainability
Mohammed Ayub and El-Sayed M. El-Alfy
ABSTRACT
Electric vehicles (EVs) are an eco-friendly mode of transportation and are transforming smart grid operations and logistics. As EV charging grows, uncoordinated home plug-ins spike peaks, overload feeders and transformers, degrade power quality, shorten equipment life, and leave network capacity poorly used. These hazards drive the use of EVs and ubiquitous smart-meter data for fine-grained EV classification, charging detection, characterization, and management. However, the classification of EV type from the charging profile is challenging due to overlapping charging patterns. Therefore, in this paper, we propose a two-stage pipeline for boosting EV type categorization directly from their charging profiles, without aggregates or forecasting, using a label-light approach and causal, deployment-ready features. In the first stage, K-Means with hysteresis is used to perform unsupervised on/off segmentation on average real power, treating it solely as data processing. In contrast, the second stage trains three gradient-boosted classifiers (HistGradientBoosting, LightGBM, and XGBoost) using charging-only samples. We evaluated the proposed models on a public dataset and compared several real-time feature sets: a minimal instantaneous set (F18-RAW), short-horizon causal statistics (F108-CAUSAL), and a two-stage variant (2F108-CAUSAL) augmented with longer univariate windows on real power (2F141-CAUSAL). Additionally, an ablation study is conducted using windowed real power (F60-SW). The results show that treating segmentation as preprocessing consistently sharpens class signatures, causal short-horizon statistics disambiguate EVs with similar instantaneous plateaus by encoding slope/stability, and modest longer-window summaries further reduce confusions in high-power overlap bands; all while remaining CPU-efficient and strictly causal with no look-ahead. The proposed pipeline turns minor, physically meaningful changes in per-EV charging behavior into strong, interpretable decision boundaries, which enable type-aware power budgeting, safe concurrent charging, and uniform interoperability tests.
Generative Adversarial Synthesis and Deep Feature Discrimination of Brain Tumor MRI Images
Md Sumon Ali and Muzammil Behzad
ABSTRACT
Compared to traditional methods, Deep Learning (DL) becomes a key technology for computer vision tasks. Synthetic data generation is an interesting use case for DL, especially in the field of medical imaging such as Magnetic Resonance Imaging (MRI). The need for this task since the original MRI data is limited. The generation of realistic medical images is completely difficult and challenging. Generative Adversarial Networks (GANs) are useful for creating synthetic medical images. In this paper, we propose a DL based methodology for creating synthetic MRI data using the Deep Convolutional Generative Adversarial Network (DC-GAN) to address the problem of limited data. We also employ a Convolutional Neural Network (CNN) classifier to classify the brain tumor using synthetic data and real MRI data. CNN is used to evaluate the quality and utility of the synthetic images. The classification result demonstrates comparable performance on real and synthetic images, which validates the effectiveness of GAN-generated images for downstream tasks.
Intelligent Self-Learning Robot for Adaptive Object Sorting
Mir Suhail Alam, Rama Khaled Alsheikh, Hunain Kaisar and Emmanuel Okafor
ABSTRACT
This study evaluates the impact of a custom value-based deep reinforcement learning agent (PQCN-MobileNetV3-Small-S-FCN-1152 architecture) on the adaptability of robotic systems in object sorting tasks. Specifically, we study performance differences resulting from training on irregular and regular blocks under dynamic conditions. As robotics becomes increasingly integrated into diverse environments, the ability to adapt to the fluctuating characteristics and configurations of objects becomes crucial. Our proposed model, which leverages a lightweight convolutional neural network framework (consisting backbone network for feature extraction and a fully convolutional neural network for value estimation), enhances the adaptability of robots when sorting objects of varying shapes and sizes. Through comparative experiments, we examine model performance metrics, focusing on accuracy and processing speed in sorting scenarios that mimic real-world variability. The results indicate that although both training modalities provide substantial improvements in terms of robotic adaptability, training with regular blocks significantly enhances the model’s robustness to unexpected changes in very dense and cluttered environmental scenarios. Additionally, the adaptability demonstrated by our proposed models under dynamic conditions suggests their potential for deployment in various operational environments. This research contributes to the understanding of optimal training approaches to improve robotic object sorting capabilities, highlighting the need for adaptable systems in the face of unpredictable real-world challenges. The results highlight the importance of innovative model architectures for advancing robotic functionality in complex tasks.
Reliability Based-model for Optimizing Routing of Truck-drone Delivery System
Ahmed Ghaithan, Awsan Mohammed, Ahmed Attia and M. N. Darghouth
ABSTRACT
This study develops a reliability-based optimization model for routing of truck-drone delivery systems, where trucks transport goods from a central warehouse to depots and drones perform last-mile deliveries. Drone failures are modeled as exponentially distributed events that may result in partial or complete loss of customer demand. The optimization model is formulated in two stages: the first stage identifies feasible routes that minimize expected demand loss, while the second stage optimizes truck depot assignments to ensure overall system reliability. A demonstrated example with two depots, three drones, two trucks, and six customers was conducted. In the first phase, 41 feasible paths were evaluated. In the second phase, three paths were selected with a minimum expected loss of 0.1999, with a total travel time of 50.21 minutes. When optimizing for travel time, only depot A was selected to serve all six customers using two drones, yielding a total travel time of 46.55 minutes and an expected loss of 0.2179. The results demonstrate that considering reliability of the delivery system significantly influence the routing decisions.
Image Registration via Corona Virus Optimizer for Visual Servoing
Inssaf Harrade, Mohamed Kmich, Hicham Karmouni, Zakaria Chalh, Mhamed Sayyouri and Mohamed Abouhawwash
ABSTRACT
Image registration plays a central role in computer vision and robotics, as it enables machines to guide their actions by interpreting visual information captured by onboard cameras. In this work, we propose a novel intensity-based image registration approach that employs the Normalized Sum of Absolute Differences (NSAD) to quantify the similarity between a current image and a reference image acquired either at an earlier time or from a different viewpoint. The NSAD metric directly measures how geometric transformations—including translation, rotation, and scaling—align the two images. Instead of relying on conventional gradient-based optimization techniques, we estimate the transformation parameters using the Corona Virus Search Optimizer (CVSO), a swarm-based metaheuristic that explores the search space until convergence of the NSAD objective. Experimental results demonstrate that the proposed CVSO-based framework is robust to noise, blur, and image perturbations, effectively avoiding local minima. Furthermore, evaluations on both synthetic and real image pairs confirm that the proposed method achieves higher accuracy and faster convergence than several state-of-the-art heuristic optimizers, producing precise alignments with reduced computational time.
A Retrieval-Augmented Generation Framework for Synthesizing Pedestrian–Autonomous Vehicle Interaction Research in Virtual Environments
ABSTRACT
The growing integration of autonomous vehicles (AVs) into urban traffic has intensified the need to understand how pedestrians perceive and respond to automated traffic environments. Virtual and extended reality (VR/XR) experiments have become essential tools for examining such interactions under safe and controlled conditions. However, different studies use different simulation platforms, experimental protocols, and analytical approaches, limiting cross-study comparability and cumulative understanding. This paper uses a retrieval-augmented generation (RAG) framework for conducting an AI-assisted literature review of immersive pedestrian–AV interaction research. A structured Scopus query identified open-access empirical studies published after 2015. Each journal paper was processed and analyzed using a multi-stage RAG pipeline that combined language-model-based extraction, semantic normalization, and network visualization. The framework harmonized heterogeneous terminology across studies and revealed methodological linkages among hardware, software, and analytical techniques. The adopted framework ensures transparency, scalability, and legal compliance in literature synthesis. Beyond summarizing current evidence on immersive pedestrian–AV research, it demonstrates how retrieval-augmented workflows can advance reproducible, data-driven reviews in transportation and human-factors domains.
GIS and Machine Learning for Identifying Crash Hotspots: Amman, Jordan as a Case Study
Duha Alsarayreh, Rana Abid, Baqer Al-Ramadan and Aws Kaffini
ABSTRACT
GIS is the most significant aid in visualizing and analyzing accidents because it easily provides a way to visualize hazardous areas on highways. The present study uses accident data with high point density based on 166,000 points collected by the Police and Security Department in Jordan. After processing the data, 106,220 accidents that occurred during 2022 were identified. Geospatial analysis conducted using network-constrained kernel density estimation indicated high densities of accidents at the north and center of Amman along major urban roads predominantly with moderate incidence in the suburbs and low frequencies in rural areas. The study area was further reclassified into polygon features to provide accident counts correlated to road segments with increased crash frequency particularly ten road segments including Hussein Khawaja Street and Al Jama’a Street where speed between 50 to 60 km/h enhance susceptibility. The Getis-Ord Gi* hot spot analysis generates an ICOUNT field and applies K-nearest neighbors with multiple comparison testing the strength of results. DBSCAN will assist by identifying and labeling clusters within the data. It is not like K-Means clustering in the sense that it can deal with noise or outliers in the data, through parameters such as eps and MinPts. The study determines the optimal epsilon using a k-distance plot, suggesting values between 800 and 1000 meters, with finer tuning between 700- and 3000-meters improving cluster performance. The analysis revealed 10 micro-risk zones, with a focus on sub-clusters 3 and 6, which had high accident rates at night with an average speed of 70 km/h. Sub-cluster 5 indicated increased female involvement during inclement weather, while sub-cluster 2 showcased daytime collisions correlated with behavior-driven risks at moderate speeds.
Towards Climate-Resilient and Low-Carbon Transport Infrastructure: A Circular Economy Perspective through Bibliometric Analysis
Zahwa Moustafa, Abdalla Moustafa and Muhammad Asif
ABSTRACT
The transport sector is a major contributor to carbon emissions and is increasingly exposed to climate-related disruptions, creating an urgent need for infrastructure that is both low-carbon and resilient. This study provides a bibliometric synthesis of Scopus-indexed journal literature (1967–2025; n = 6,367) to map how circular economy (CE) research intersects with decarbonization and climate resilience in transport infrastructure. Using VOSviewer science-mapping, we analyze publication and citation evolution, leading outlets and countries, and thematic structures based on keyword co-occurrence and term clustering. Results show accelerated growth after 2015, reflecting alignment with global sustainability agendas and the rapid expansion of CE-related transport research. The thematic landscape is organized around four interlinked domains: (i) technological innovation and system optimization, (ii) energy, emissions, and environmental assessment, (iii) infrastructure durability and material performance, and (iv) socio-behavioral and policy dimensions. Cross-country patterns indicate that some nations achieve high impact with lower volume due to concentrated contributions in highly cited, policy-relevant research and strong international collaboration networks. Overall, the study consolidates research streams that are often examined separately and identifies actionable directions for policy design, infrastructure planning, and future research on circular, low-carbon, and climate-resilient transport systems.
Recent Innovations in Last-Mile Delivery: Drones, Autonomous Vehicles, and Beyond
Rahaf Alqadda, Wahad Alshmlani and Irfanullah Khan
ABSTRACT
Last-mile delivery has undergone significant changes due to technological advancements and the rise of e-commerce. Companies must explore innovative technologies to meet evolving customer expectations for faster, more efficient, and cost-effective delivery solutions. Technological innovations like drones, autonomous ground vehicles, and advanced delivery mechanisms have emerged rapidly. Drones bypass traditional road traffic, enabling quicker deliveries, especially in congested areas and remote locations. Autonomous ground vehicles navigate streets with minimal human intervention, reducing labor costs and improving delivery precision. This paper reviews last-mile delivery, focusing on the transformative role of drones, the operational efficiency of autonomous ground vehicles, and future innovations that will revolutionize the industry. It examines the current landscape, technological advancements, regulatory hurdles, and potential growth areas to provide a holistic understanding of how last-mile delivery is evolving in response to modern marketplace demands.
A Framework for Integrating Human Factors into Industry 4.0 Maintenance
Aamir Hussain and Ahmet Kolus
ABSTRACT
This study discusses the integration of human factors in the maintenance process of Industry 4.0 by focusing on the improvement of system efficiency, reliability, and human well-being. A literature-based analysis outlines four categories of factors, including individual, technological, process, and organizational factors, which together influence human-system interaction and performance. The proposed framework emphasizes the need for training, ergonomics, predictive maintenance, data management, and leadership support as key factors in human-centered maintenance. Findings show that technological and process dimensions have been widely explored, while human and organizational dimensions are relatively less explored. The suggested framework provides practical guidance for making human capabilities and digital technologies compatible to enhance productivity, safety, and sustainability in Industry 4.0 environments.
System Identification of Dynamic Adaptive Fuzzy Modeling Using Reinforcement Learning
Mohamed Soliman, Mohammed Abdel-Nasser, Ahmed Abubaker and Mahmoud Abdelaal
ABSTRACT
Dynamic Adaptive Fuzzy Modeling (DAFM) is a flexible framework for describing nonlinear, time-varying systems, yet its parameters must adapt reliably to changing conditions. We propose a reinforcement-learning (RL) approach that performs system identification online by embedding a temporal-difference learning rule with eligibility traces into the DAFM update law. The method updates membership and consequent parameters directly from streaming data, assigning greater credit to recent errors while remaining interpretable through the fuzzy rule base. We validate the algorithm on two nonlinear numerical examples and a laboratory liquid-level rig. Across these studies, the RL-driven DAFM achieves real-time identification with accurate tracking and strong robustness to nonstationary dynamics, demonstrating a practical route to data-efficient, interpretable, and adaptive fuzzy modeling.
Blockchain Enabled Traceability System for Verifiable Halal Meat: Enhancing Transparency and Trust in Food Logistics
Mujtaba Zaman, Eslam Alanazy and Irfanullah Khan
ABSTRACT
Halal verifiability in the food industry remains a challenge. Many factories have periodic audits of their factories to make sure standards are upheld. However, these are snapshots of the facilities as audits are not taking day to day, in addition to the audits being of a subjective nature. This lack of transparency can lead to consumer trust falling in meat processing plants. To address this gap in lack of consumer trust, we propose a blockchain-enabled meat traceability system that tracks different parts of meat production in meat processing plants. This will improve transparency and trust as this system can be automated using IOT and Smart Contracts to receive verification of actions that have been done correctly and shown to end consumers so he/she can make their own decision. We made a functional prototype for looking at the feasibility of the blockchain system. This study proposed the use of a website and mobile application to allow access to the blockchain. The mobile application also allows QR code scanning for easy looking up records in blockchain. In blockchain, hashing was used for implementing data integrity and immutability of records. The study also demonstrates how religious requirements can be implemented using Smart Contracts and IOT devices to achieve automated results based on actions like phrase reciting without human intervention. Implementing blockchain, Smart Contracts, and IOT at a large scale do have some challenges like scalability, data storage, privacy, and governance. The prototype created shows the technical feasibility of the project requiring regulatory collaborations from all stakeholders to become a transparent consumer powered meat verification system.
Intrusion Detection 2.0: Revisiting Benchmark Workflows with Explainable and Tuned AI Models
F M Jahiduzzaman, Md. Aqib Aman, Md Moazzem Hossain and Ashiqur Rahman Ashiq
ABSTRACT
Intrusion detection systems (IDSs) are a cornerstone of network security, yet designing detectors that generalize across diverse attack types remains challenging. This work reproduces the end-to-end workflow proposed by Al-Doori and Alheeti (2025) for “AI Driven Features for Intrusion Detection and Prevention Using Random Forest.” The original paper combines feature optimization (Genetic Algorithm and Particle Swarm Optimization) with Random Forest classifiers and Grey Wolf Optimization, achieving an accuracy of 0.94, precision 0.95, recall 0.93, and F1 score 0.94 on the NSL-KDD dataset. We implement a similar pipeline on the KDDCup’99 (10%) dataset, perform z-score normalization and one-hot encoding, then compare a compact deep neural network (DNN) against CatBoost, LightGBM, XGBoost, and TabNet classifiers. Our best model (CatBoost) achieved an overall accuracy of 0.999 with a macro-F1 of approximately 0.94 and a weighted-F1 of approximately 0.999 on the held-out test set. We also report precision 0.985, recall 0.987, and F1 0.985 for the baseline DNN model, substantially exceeding the anchor paper’s scores due to dataset differences and model capacity. We analyze per-class performance and observe that rare attack types remain harder to detect, a finding consistent with recent IDS literature. All methods and experiments are described to facilitate reproducibility and provide a comprehensive comparison to the anchor study.
Reinforcement Learning based Quantum Approximate Optimization Algorithm for Vehicle Routing Problem
Md. Aqib Aman and Fardin Azam Sakib
ABSTRACT
Reinforcement Learning–based Quantum Approximate Optimization Algorithm (QAOA-RL) is used to solve the Vehicle Routing Problem (VRP) by transforming it into a Quadratic Unconstrained Binary Optimization (QUBO) form. The RL agent learns to optimize QAOA parameters for better route selection. Results show that QAOA-RL achieves a lower total cost compared to exact ILP and heuristic Greedy Search methods. Although the training time is higher, the solution quality improves significantly for larger problem sizes. This study highlights QAOA-RL as a scalable and promising hybrid quantum–classical approach for real-world logistics optimization.
Offline IoT-Enhanced Open Vehicle Routing for Perishable Products: A Resilient Optimization Framework
Saphiah Bajba and Ahmed Attia
ABSTRACT
This paper presents an offline IoT-enhanced optimization framework for the Open Vehicle Routing Problem with Time Windows (OVRPTW) in perishable goods logistics. The proposed model leverages historical IoT and GPS data, including temperature logs,travel times, and failure records, to pre-plan cost-efficient and resilient delivery routes. The Mixed-Integer Linear Programming (MILP) formulation integrates perishability constraints, heterogeneous multi-compartment fleets, driver-workload equity, and resilience mechanisms through the use of backup vehicle assignments. The model was implemented using PuLP–CBC and tested on a cold-chain instance. Numerical results confirm that the framework minimizes freshness-loss penalties and operational costs while maintaining service reliability and workload balance. A sensitivity analysis reveals that perishability has the most significant influence on total cost, followed by fleet and equity factors. In contrast, variations in travel time have a minor effect.
Performance Evaluation of Evolutionary and Swarm Algorithms for UAV Attitude Control
Md. Hafizur Rahman and Muhammad Gulzar
ABSTRACT
Unmanned Aerial Vehicles (UAVs) require precise control over their pitch and roll angles to maintain stability and maneuverability. Traditional PID controllers are often used for this purpose; however, manual tuning of PID parameters can be time-consuming and suboptimal. This paper presents a comparative study of four optimization techniques; Grey Wolf Optimizer (GWO), Ant Colony Optimization (ACO), Genetic Algorithm (GA), and Particle Swarm Optimization (PSO) for automatic tuning of PID parameters in the pitch/roll control system of a UAV. Simulation results demonstrate the efficacy of each technique in improving system performance, measured by overshoot, settling time, and mean absolute error. Based on the Overshoot, Peak Time, and Rise Time, both PSO and GA performed better in terms of faster and more stable system responses.
Optimal Control of Active Quarter-Car Suspensions Using PSO/TLBO-based Optimization and LQR/LQG Strategies under ISO Road Excitations
Sadiq Ibrahim Shuaibu, Muhammad Gulzar, Salman Habib and Khaleel Mohamed
ABSTRACT
This paper presents a systematic comparison of PID, LQR, and LQG controllers for active quarter-car suspension, focusing on balancing ride comfort, handling stability, and control effort. PID gains and LQR weighting matrices are optimized using Particle Swarm Optimization (PSO) and TeachingLearning-Based Optimization (TLBO) algorithms. Performance is evaluated on half-sine bump and ISO 8608 random road profiles, with key metrics including RMS body acceleration, suspension and tire deflections, actuator effort, and transient response features such as overshoot and settling time. Statistical analysis over 20 randomized runs reveals that LQR controllers tuned via TLBO and PSO maintain ride comfort comparable to passive suspension while significantly reducing suspension deflection and control effort. PID controllers achieve moderate improvements but incur higher actuator effort and suspension travel. LQG offers minor gains over passive suspension but does not outperform optimized LQR controllers. These findings provide practical guidance for designing active suspension controllers that efficiently trade off comfort, stability,
|
| 09:40 AM - 10:20 AM |
Session 14: Saudi Railways (SAR) |
| 10:20 AM - 10:40 AM |
Refreshments and Networking |
| 10:20 AM - 11:10 AM |
Session 15: Keynote-III |
| 11:10 AM - 11:40 AM |
Session 16: Keynote-IV |
| 11:40 AM - 01:00 PM |
Prayer and Lunch Break |
| 01:00 PM - 02:00 PM |
Session 17: Industry Engagement |
| 02:00 PM - 04:30 PM |
Session 18: QCAR Competition |
| 02:00 PM - 02:40 PM |
Session 19: Panel Discussion II: Industry Engagement in Mobility and Logistics |
| 02:40 PM - 02:50 PM |
Refreshments and Networking |
| 02:50 AM - 04:30 AM |
Session 20A: Vehicle Systems, Electrification & Charging (Track: SRTS)
Analytical and Machine Learning–Based Decision Support for Autonomous EV Charging Station Selection
02:50 PM
Ahmed Attia, Mohammed Algafri, Mohammad Abdel-Aal and Anas Alghazi
ABSTRACT
This study develops a decision-support framework that combines analytical and machine learning methods for selecting charging stations (CSs) for autonomous electric vehicles (EVs). Stations are evaluated using five criteria: travel time, waiting time, charging duration, energy price, and required energy. The framework applies the Fuzzy Analytical Hierarchy Process (FAHP) with the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS), Grey Relational Analysis (GRA), and K-Means++ to estimate weights and rank alternatives. A case study of six CSs shows that subjective (FAHP), objective (GRA), and data-driven (K-Means++) weighting produce different rankings: FAHP favors user preferences, GRA stresses efficiency under incomplete data, and K-Means++ offers scalable insights for large datasets. Together, these methods demonstrate complementary strengths and provide a practical basis for real-time EV-CS selection in smart cities.
Examine the E-Bike and E-Scooter User Behavior and Optimizing Charging Locations for Sustainable Campus
03:10 PM
Renada Abd Alkader, Noura Alowayed, Noman Sahito and Latifah Ghorab
ABSTRACT
This study examines the e-bike and e-scooter user behavior and optimizes charging locations to increase on-campus mobility and achieve sustainable development in King Fahd University of Petroleum and Minerals (KFUPM). To achieve this aim, our study used a mixed-methods approach: a parametric simulation model in Rhino Grasshopper was developed to identify optimal layouts for e-bike and e-scooter charging stations, and a questionnaire was designed to collect data on student user behavior. This user-preference data was converted into spatial coordinates for 29 station placements. Mostly, students need charging locations around certain academic buildings, bus stations, housing, and mall. The questionnaire results show that students are willing to use e-bikes and e-scooters if a proper charging location is available with a shared e-bike system at a subsidized rate. The study suggests that the subsidized sharing bike system may introduce a shaded charging station to enhance the on-campus micro mobility concept.
Accurate battery parameters estimation using Newton-Raphson-based optimizer
03:30 PM
Mahmoud S. Abouomar, Walid Merrouche, Nur Hamida, Md Samiullah and Sami El Ferik
ABSTRACT
Precise estimation of the equivalent circuit model (ECM) parameters is essential for improving the predictive capability of lithium-ion battery (LiB) models and enhancing the performance of battery management systems (BMSs) in electric vehicles. This study proposes an artificial intelligence (AI)-driven methodology for ECM parameter identification based on the Newton-Raphson-Based Optimizer (NRBO) algorithm. The proposed method minimizes the deviation between simulated and experimental terminal voltages, achieving high estimation accuracy. The NRBO’s effectiveness is validated using real-world data and benchmarked against state-of-the-art metaheuristic algorithms, including the Sine Cosine Algorithm (SCA), Harris Hawks Optimization (HHO), and Salp Swarm Algorithm (SSA). Comparative results demonstrate that NRBO provides superior accuracy, underscoring its potential as an efficient and robust optimization framework for LiB parameter estimation and advanced BMS applications.
Predefined-Time Virtual Synchronous Generator Control for Voltage Stability Enhancement in DC Microgrids using Hybrid Electric Vehicle Support
03:50 PM
Hossam Ali, Mohamed Zaery and Mohammad A. Abido
ABSTRACT
The increasing penetration of renewables has driven the evolution of converter-dominated DC microgrids (MGs), which offer improved efficiency and simplified control architectures compared to conventional AC systems. However, the absence of mechanical inertia in converter-based DC MGs results in pronounced voltage instability during sudden load or generation fluctuations. To address this challenge, this paper proposes a predefined-time virtual synchronous generator (PdT-VSG) control strategy integrated with a hybrid electric vehicle (HEV) as an energy storage system. The HEV system integrates a fuel cell, a supercapacitor, and a battery to provide coordinated virtual damping and inertial support. Irrespective of the initial operating conditions, the predefined-time controller is intended to guarantee that the DC-bus voltage returns to its nominal level within a specified time bound. Simultaneously, the VSG emulates the behavior of a conventional synchronous machine, thereby enhancing system damping and transient stability. To assess the controller's performance under different load situations and low-inertia scenarios, a comprehensive MATLAB/Simulink model of an islanded DC MG was developed. The outcomes demonstrate that the PdT-VSG-based HEV provides superior transient voltage recovery, less overshoot, and a quicker settling response compared with existing inertia-emulation and damping-based control approaches. Overall, the proposed control scheme offers a fast, predictable, and robust voltage regulation solution well-suited for future hybrid and renewable-dominated DC MGs.
Improving Ride Comfort in Autonomous Vehicles with Fractional-Order SMC Active Suspension under Road Uncertainties
04:10 PM
Khaled Bin Gaufan, Abdulrazaq Nafiu Abubakar, Mubarak Badamasi Aremu, Jamilu Umar Yahaya and Nezar Alyazidi
ABSTRACT
Ensuring ride comfort, handling stability, and road adaptability is crucial for developing smart and resilient transportation systems. This paper presents a novel Fractional-Order Sliding Mode Control (FOSMC) strategy for active suspension systems in Connected and Autonomous Vehicles (CAVs/AVs), aiming to enhance ride comfort and robustness under varying road and load conditions. Unlike conventional integer-order SMC approaches, the proposed FOSMC integrates fractional calculus into the sliding surface design, introducing memory and hereditary dynamics that more accurately represent vehicle behavior. This framework mitigates the effect of disturbances, reduces chattering, and enhances actuator operation for intelligent and efficient mobility solutions. The controller’s stability is rigorously analyzed using Lyapunov theory, confirming global asymptotic stability. Simulation results demonstrate that the proposed FOSMC reduces suspension deflection by 36\%, and control chattering by more than 50\% when compared to traditional SMC, demonstrating its potential to improve comfort, stability, and safety in intelligent mobility applications.
|
| 03:50 PM - 04:30 PM |
Session 20B: Mathematical Optimization & Algorithmic Decision Models in Logistics (Track: AOIL)
An Adaptive Simulation-Optimisation Framework for Multi-Echelon Humanitarian Relief Networks
03:50 PM
Ismail Mohamed Ali, Reda Ghanem, Sara Salim, Sondoss Elsawah, Li Qiao and Yu Zhang
ABSTRACT
Humanitarian supply chains (HSCs) operate under extreme uncertainty, where rapid and informed decision-making is essential to ensure efficient relief delivery. This paper presents a simulation–optimisation framework that integrates a discrete-event simulation (DES) model with a multi-objective evolutionary algorithm (NSGA-II) to jointly optimise workforce allocation across distribution nodes and routing decisions within a multi-echelon relief network. The framework aims to maximise demand satisfaction while minimising total operating cost. It was validated through two experimental configurations: (i) a pre-defined routing experiment, where node-to-node routes were fixed, and (ii) an adaptive routing experiment, where routes were treated as decision variables within the optimisation process. The results show that the adaptive routing configuration expanded the Pareto front by 132.1% and increased mean demand satisfaction by 57.1% relative to the fixed-route baseline. These findings highlight the framework’s optimisation efficiency and adaptability in generating diverse, high-quality non-dominated solutions that effectively balance operational costs and demand satisfaction, demonstrating its value as a robust decision-support tool for humanitarian logistics under uncertainty.
Risk-Based Optimization of Driver Allocation in Urban Last-Mile Delivery
03:10 PM
Usman Ibrahim, Samhar Alzghaier and Mohammad Aldurgam
ABSTRACT
Uncertainty in last-mile delivery complicates day ahead driver staffing. We propose a single-period stochastic optimization model that jointly chooses base (pre-booked) and surge (backup) capacity. The model combines the newsvendor overage / underage tradeoff, dual-sourcing cost differentiation, and exogenously specified risk-based route classes to represent heterogeneous demand. Given counts of high, medium, and low risk routes and unit costs, it computes the cost with the objective of finding the optimal numbers of pre-booked and backup drivers. An illustrative example shows how risk segmentation and dual sourcing shape the optimal mix and quantify the tradeoff between higher upfront staffing and lower shortfall exposure. We use finite convolution formulas for the distribution of total driver demand under independent per route needs, enabling efficient evaluation of expected penalties across candidate staffing plans. The model can be extended to rolling horizon and multi-period settings.
A Novel Adaptive Large Neighborhood Search Heuristic for the UAV Routing Problem with Multiple Time Windows
03:30 PM
Omayma Sedik and Slim Belhaiza
ABSTRACT
Unmanned Aerial Vehicles (UAVs) have emerged as a promising alternative for last-mile delivery, offering cost-effective, rapid, and flexible logistics solutions. However, their limited battery life, payload capacity, and range impose significant operational challenges. This paper presents a novel Adaptive Large Neighborhood Search (ALNS) heuristic to solve the UAV Routing Problem with Multiple Time Windows (UAV-RPMTW). The proposed ALNS framework dynamically selects destroy-and-repair operators based on adaptive performance feedback. The hybrid structure efficiently balances exploration and exploitation through diversified search operators and tabu-based restart strategies. Extensive computational experiments on newly designed UAV-RPMTW benchmark instances demonstrate that the proposed approach achieves high-quality solutions with improved computational efficiency. The results confirm that allowing multiple time windows enhances both service flexibility and cost efficiency in UAV delivery systems.
An efficient decoding algorithm for the Node-Shift Encoding representation
03:50 PM
Souad Abdoune and Menouar Boulif
ABSTRACT
The Node Shift Encoding (NSE) for genetic algorithms ensures feasibility in permutation problems like the Traveling Salesman Problem, a cornerstone in logistics optimization. However, NSE suffers from high decoding complexity, due to nested rank updates. We propose an optimized decoder that replaces rank tracking with a direct shift-and-insert mechanism using a position mapping table. Experiments on instances ranging from 100 to 100000 cities show a 2–3× speedup over the original NSE decoder, without increasing memory consumption. While both algorithms share a O(n2) worst-case theoretical complexity, the proposed approach reduces practical runtime, making it effective in handling large-scale problems.
A Smart Multi-Objective Optimization Framework for Sustainable Supplier Selection in Renewable Energy Equipment Supply Chain
ABSTRACT
The renewable energy sector faces critical challenges in establishing reliable supply chains for essential equipment components. This research develops an integrated decision support framework for supplier evaluation and order distribution in solar panel component procurement. Unlike existing approaches, this study simultaneously optimizes three fundamental business dimensions: economic efficiency, product reliability, and delivery performance, while incorporating regulatory compliance verification through government-certified testing facilities. The proposed multi-period multi-objective renewable energy supply chain management model addresses the practical needs of emerging renewable energy enterprises that lack comprehensive testing infrastructure. A hybrid solution methodology combining interactive fuzzy programming with weighted goal programming is employed to handle inherent uncertainties in supplier performance data. The framework integrates expert committee judgments weighted by individual experience levels to prioritize conflicting objectives. Computational experiments demonstrate achievement of 48.32% satisfaction for cost minimization, 77.89% for quality maximization, and 93.67% for time reduction objectives. Sensitivity analyses confirm the model's robustness under varying decision-maker preferences and uncertainty levels. This research contributes a comprehensive decision framework for renewable energy supply chain management, offering direct applicability to industries requiring regulatory oversight in supplier qualification processes.
|
| 02:50 PM - 04:30 PM |
Session 20C: Decarbonization & Energy Transitions in Cognitive Cities (Track: CCSTE)
Muhammad Aamir Basheer
02:50 PM
Determinants of First and Last Mile Mode Choice to Transit Stations
ABSTRACT
The first and last mile (F&LM) are crucial aspects of transit services shaping the experience of transit users. Understanding the factors influencing travellers’ choice of first and last mile is essential for improving multimodal connectivity and promoting sustainable mobility. This study analyzes F&LM mode choice using survey responses on how travellers’ access/egress transit stations. A multinomial logit (MNL) model was estimated in SPSS using a forward stepwise approach. F&LM mode choice was estimated using socio-demographic factors (age, gender, education, and income) alongside trip purpose and standardized travel time and distance. Distance to the transit station emerged as the most influential determinant of non-walking choices, followed by education level and trip purpose. Shopping and social/recreational trips showed the highest likelihood of using Car/Taxi/Rickshaw, while education-related trips were more likely to involve motorcycles. The findings highlight the critical role of spatial accessibility and socio-economic differences in shaping F&LM travel behaviour and suggest that station-area planning should prioritize walkability, safety, and equitable design for diverse trip purposes and traveller groups.
Exploring the Potential of Renewable Energy for Sustainable Mobility: A Simulation- Based study of Hydrogen Vehicle Penetration in Oman’s Road Network
03:10 PM
Siham Farrag, Tarek Sheltami, Fatma Outay and Ansar Yasar
ABSTRACT
Hydrogen fuel is gaining attention as a promising zero-emission energy source, aligning with global sustainability goals and supporting the transition to zero carbon emissions. This study examines the potential of using hydrogen as an alternative fuel for a sustainable mobility in Muscat, Oman. We developed an integrated modelling framework that combines microscopic traffic simulation, energy demand modeling, refueling infrastructure station’ estimation, and well-to-wheel (WTW) emissions evaluation. A microscopic simulation software (SUMO) was applied to evaluate the penetration rate of hydrogen-powered vehicles (0%, 20%, 40%, 60%) with different hydrogen production pathways. Results indicate that with a 60% penetration of green hydrogen, total greenhouse gas (GHE) emissions can be reduced by nearly 95%. Blue hydrogen achieved intermediate benefits by about 60% reduction, whereas grey hydrogen increased total emissions due to fossil- based feedstock. Moreover, cumulative hydrogen consumption increases nonlinearly with hydrogen vehicle penetration level. The required number of refueling stations increases exponentially due to quantity and temporal demand. These findings provide decision-makers with quantitative perception into hydrogen adoption, infrastructure requirements, and emissions mitigation pathways
Zero-Carbon Adsorption Cooling System Optimization for Maritime Transport Using MOF-303
03:30 PM
Rached Ben-Mansour
ABSTRACT
Global warming is already producing serious and wide-ranging impacts on climate systems, ecosystems, and human societies, with many changes expected to persist for centuries. Human activities remain the dominant driver of these effects, and urgent measures are required to mitigate their consequences. Maritime transportation, which currently carries nearly 80% of global goods, is among the most energy-efficient modes of transport, yet further improvements are needed to reduce its carbon footprint. Cooling demand alone accounts for 10–30% of total shipboard energy consumption, particularly for passenger spaces, refrigerated cargo, and food products. This study proposes a zero-carbon adsorption cooling system powered by waste heat represented in exhaust gases and seawater for cooling, thereby addressing both environmental sustainability and operational efficiency. A dynamic model was developed to investigate the performance of an adsorption cycle employing MOF-303, a highly water-selective metal–organic framework with superior uptake capacity and kinetics compared with conventional silica gel or zeolite. The system operates through alternating desorption, condensation, evaporation, and adsorption processes, simultaneously providing cooling and freshwater production. Parametric studies examined the influence of half-cycle duration, driving hot-water temperature, and sorbent mass on coefficient of performance (COP), specific cooling power (SCP), and cooling capacity. Results indicate that short cycle times maximize throughput but limit efficiency, whereas long cycles enhance COP and freshwater yield at the expense of cooling power. An optimal operating window was identified at intermediate cycle times (500-700 s), delivering a maximum SCP of 421 W/kg and a cooling capacity of 16.8 kW. Similarly, hot-water inlet temperatures of 85–90 °C and sorbent inventories of 20–23 kg provided the best compromise between efficiency, compactness, and output. The findings confirm that MOF-303 adsorption systems can effectively harness marine waste-heat streams to deliver sustainable cooling, offering a practical pathway toward zero-carbon maritime transport.
Modelling Transport Energy Futures Using the Low Emissions Analysis Platform
03:50 PM
Muhammad Shahid, Amjad Ali, Sikandar Abdul Qadir and Md Tasbirul Islam
ABSTRACT
Pakistan’s transport sector is a rapidly growing source of final energy demand and greenhouse-gas emissions. This study applies the Low Emissions Analysis Platform (LEAP) to assess energy-use pathways under three transport policy scenarios from 2020 to 2040. A Reference Scenario (RS) representing current policies, an Electric Vehicle Scenario (EVS) involving progressive substitution of liquid fuels with electricity across road, rail, air transport, and an Efficient Combustion Scenario (ECS) emphasising fuel economy improvements in conventional internal-combustion vehicles. Results indicate that total transport energy demand in the RS rises from 15.8 Mtoe in 2020 to 38.1 Mtoe by 2040, driven by growing fuel use in road freight and passenger transport. The EVS shifts the fuel mix towards electricity, which reaches 9.4 Mtoe in 2040, while gasoline and diesel fall to 11.6 Mtoe and 8.5 Mtoe, respectively. Although the EVS primarily represents fuel substitution, however its total final energy (30.2 Mtoe) is lower than RS and nearly equal to ECS (30.1 Mtoe) as electric drivetrains require less final energy to deliver the same transport output. The ECS shows a total energy demand about 21% lower than the RS, mainly due to improvements in combustion efficiency with gasoline and diesel declining to 14.7 Mtoe and 12.9 Mtoe, respectively. Over the modelling period, cumulative energy consumption is 533.2 Mtoe (RS), 470.4 Mtoe (EVS), and 468.6 Mtoe (ECS). These findings indicate that while the EVS pathway mainly alters the energy mix through electrification, it also realises efficiency-linked demand moderation. The ECS pathway delivers comparable reductions through improved combustion and logistics efficiency. Together, they highlight complementary approaches for reducing fossil fuel dependence and guiding Pakistan’s sustainable transport-energy transition.
Exploring the Feasibility of Adopting the Chinese Electric Mobility Framework in Saudi Arabia
04:10 PM
Muhammad Ahmad, Sheeraz Iqbal, Md Shafiullah, Atif Saeed Alzahrani and Salman Habib
ABSTRACT
One of the critical milestones under Saudi Arabia's Vision 2030 is to achieve net-zero emissions by the end of 2060. Under the prevailing situation, CO2 emissions from combustible fuels totaled 534.68 MtCO2 in 2023, accounting for a 1.5% share in global emissions, from the Kingdom. Additionally, it was noted that 27% of total emissions are contributed by the transportation sector only. To mitigate this situation, from policy to practice, several measures are essential to achieve the ambitious targets of net-zero emissions. This paper takes a significant market player in the e-mobility sector, i.e, China, as a benchmark to evaluate the impact of adopting the e-mobility framework in the Saudi Arabian context. Replicability is analyzed using qualitative and quantitative data within the scope of multi-criteria decision analysis (MCDA). Key factors such as policy, governmental support, technology, enforcement, and impact analysis under the Chinese e-mobility model are explored. Subsequently, based on the benchmark analysis under the MCDA approach, the feasibility of replicating the model in the Saudi Arabian context is investigated. The results of the MCDA, based on qualitative and quantitative analyses, indicate that replicating the Chinese e-mobility model in Saudi Arabia will remain a notable challenge, given domestic constraints across technical and non-technical areas.
|
| 02:50 PM - 04:30 PM |
Session 20D: AI-Enhanced Autonomy, Collaboration, and Decision Making (Track: CHTML)
Teaching-Learning-Based Optimization for Robust and Adaptive Control of Mechanical Ventilation Systems
02:50 PM
Gamal Muneer, Muhammad Gulzar, Salman Habib and Adnan Shakoor
ABSTRACT
Mechanical ventilation is vital for patients with respiratory disorders, requiring accurate airflow control to adapt to changing lung conditions. PID controllers are widely used, but traditional tuning methods are time-consuming and often inadequate under disturbances. This study compares five metaheuristic optimization techniques—TeachingLearning-Based Optimization (TLBO), Genetic Algorithm (GA), Ant Colony Optimization (ACO), Particle Swarm Optimization (PSO), and Grey Wolf Optimization (GWO), for automatic PID tuning for the respiratory system model under normal and obstructive conditions. Evaluation using MSE, MAE, and R² shows that TLBO achieved the best performance, while GA, PSO, ACO and GWO offered higher frequency oscillations. The TLBO-optimized PID demonstrated robustness against sudden disturbances, highlighting the potential of metaheuristic tuning to improve ventilator adaptability and efficienc
LLM-Guided Human–Drone Interaction for Autonomous Mission Execution
03:10 PM
Malak Alsufaian, Maisa Alghauainam and Hussein Samma
ABSTRACT
Recent advances in language and vision models are redefining how humans interact with autonomous systems. This paper presents an intelligent framework that exemplifies cognitive human-machine collaboration in indoor mobility applications. The system interprets spoken or written commands in natural language and executes corresponding aerial missions by combining speech recognition, large language models, and vision-based reasoning. Through this integration, the drone can understand human intent, analyze its environment, and perform context-aware actions such as identifying people, recognizing sensitive information, or auditing workstation screens. A web-based interface enables real-time interaction and feedback, supporting seamless cooperation between the operator and the drone. The framework was evaluated in several real-world scenarios using a lightweight DRONE platform, demonstrating strong adaptability and consistent decision-making. The results illustrate how multimodal language and vision understanding can enhance autonomy, transparency, and safety in collaborative aerial systems, aligning with emerging directions in cognitive mobility and logistics.
Examining the Joint Impact of Digital Dexterity and Collaboration on Supply Chain Performance and Quality Management
03:50 PM
Shama Khan, Muhammad Tayyab, Hira Tahir and Irfanullah Khan
ABSTRACT
This paper explores the effect of digital dexterity (DD) and supply chain collaboration (SCC) on supply chain quality management (SCQM) and supply chain performance (SCP). It further examines how supply chain relational capital (SCRC) moderates SCQM-SCP relationship. The theoretical foundation of the research lies in the Dynamic Capabilities View (DCV), Relational View (RV) and the Social Network Theory (SNT). We combine the technological, relational and quality-management connections in explaining how organisations could develop resilient, quality-focused and digitally agile supply chains. An instrument of a quantitative design that included a structured questionnaire was utilized to collect data from professionals in the petrochemical, transport, and manufacturing industries. Sixty non-coded responses were tested with hypothesis tests using a Partial Least Squares Structural Equation Modeling (PLS-SEM) with 5,000 (sample) bootstrapping to test hypotheses and determine reliability, validity and the significance of proposed paths. Findings indicate that digital dexterity (b = 0.395, p < 0.001) and collaboration (b = 0.332, p = 0.012) have significant influence on SCQM which is highly related to SCP (b = 0.535, p = 0.001). The SCP is directly enhanced by the relational capital (b = 0.275, p = 0.039), yet the relationship between SCQM and SCP is not mediated by the former (b = -0.055, p = 0.346). The mediation analysis supports the hypothesis of partial mediation of DD-SCP path (b = 0.212, p = 0.013) and no mediation of SCC-SCP (b = 0.178, p = 0.052). The model covers a major part of SCP variance of 62.2% which indicates its high explanation ability. The model results complement DCV and RV by showing that digital dexterity and collaboration are complementary antecedents of SCQM, which is translated into a better performance.
Closing the Transparency Gap: Evaluating Explanation Timing and Complexity via Policy Alignment in HRI
ABSTRACT
The effectiveness of explanations in Human-Robot Interaction (HRI) is often gauged through subjective measures like trust, lacking robust quantitative metrics for mental model alignment. This paper introduces Policy Divergence (Dₚₒₗ), a novel metric based on the Kullback-Leibler divergence between a robot's true policy and a human's inferred policy, to quantitatively assess explanation effectiveness. A simulation framework leveraging a Bayesian Theory of Mind (BToM) is proposed to model how humans update their beliefs based on robot explanations within a grid-world navigation task. Systematically evaluating the factors of timing (pre-action vs. post-action) and complexity (low, medium, high), the results show that pre-action explanations significantly outperform post-action in reducing policy divergence and improving predictive accuracy. High-complexity explanations yield the strongest alignment but introduce a trade-off with cognitive load. The results validate (Dₚₒₗ) as a principled metric and demonstrate that optimal explanation strategies are contingent on a critical balance between timing, detail, and the human's cognitive state.
A Reproducible Simulation Workflow for ISO-Inspired Hybrid Wheeled-Legged Wheelchair Using URDF and PyBullet
04:10 PM
Mark Abdelmessih, Tarek Eissa and Ammar Ayad Alzaydi
ABSTRACT
Equitable access to the built environment requires assistive mobility devices that can be developed and assessed systematically before clinical deployment. A research platform for a hybrid, multi-leg wheelchair mechanism is introduced to advance this goal. The contribution comprises a proof-of-concept computer-aided design (CAD) model relying on international organization for standardization (ISO) standards, a corresponding unified robot description format (URDF) model suitable for physics-based simulators, and a set of standardized geometric test layouts inspired by ISO 7176-5. The mechanism is conceived to operate in rolling and leg-assisted modes so that everyday obstacles—such as curbs, stair steps, gaps, snow, sand, and uneven outdoor surfaces—can be addressed within a unified framework. The standardized layouts (including corridor width, doorway entry depth, right-angle turns, and reversing width) provide a common context for assessing maneuverability and minimum turning radii while enabling fair comparison of future controllers. Representative simulations confirm successful model import and commanded motion across intended ranges without prescribing specific control strategies. By lowering the barrier to rigorous, reproducible testing in safe virtual settings, the platform is positioned to accelerate research toward more capable mobility aids, with the potential to enhance independence, dignity, and participation for people with disabilities.
|
