Abdellah Chehri

Automating the extraction of domain models from natural language requirements remains a central challenge in model‐driven engineering (MDE), particularly at the level of the computation‐independent model (CIM), the most abstract layer of the model‐driven architecture (MDA). This paper introduces an approach that uses large language models (LLMs) to generate domain models from software requirements. Unlike prior work that relies on a single monolithic prompt or on generic task decomposition, our approach organises the modelling task as a modular chain of prompts with dedicated stages for enumerations, entities, and relationships, each producing an explicit intermediate output that feeds the next stage before a final consolidation step. The approach is implemented as a Visual Studio Code extension called SmartModel, which integrates generation, visualisation, and requirement‐level traceability into the developer's environment. To evaluate the approach, a case study on smart tourism software requirements was conducted, comparing the LLM‐generated domain model with an expert‐constructed reference model using both qualitative expert assessments and quantitative metrics. Results show that LLMs achieve high precision in identifying entities and enumerations, whereas relationship and attribute extraction remain considerably less reliable and require systematic human review. The findings underscore the value of LLMs in accelerating early‐stage MDE tasks, while highlighting the necessity of human feedback to ensure semantic integrity. This work contributes a practical and extensible tool and empirical insights that inform future research into hybrid AI‐human modelling workflows and advanced LLM techniques for software modelling.

Driving assistance and autonomous driving are among the fastest-evolving domains of intelligent transportation systems (ITSs). However, visually impaired pedestrians (VIPs) remain weakly protected by roadsides or vehicle perception systems, especially when the right of way must be communicated in an accessible and auditable manner. In this paper, we describe a safety-constrained vehicle-to-pedestrian (V2P) architecture designed for VIP crossing assistance. The system creates a time-bounded interaction state between pedestrians, roadside infrastructure, and vehicles at the Fog level to avoid permissive instructions before all the safety conditions are satisfied. Our objective is not to design a new detector, tracker, or MLLM model. Instead, we combine edge tier descriptors, Fog-tier multi-view descriptors, and multi-camera multiple-object tracking (MC-MOT) within a latency-constrained V2P loop. Our main empirical focus is WildTrack multi-camera continuity with novel multi-camera consensus gating (MCCG), a geometric-support safety gate. The reproduced Fog-tier continuity branch reaches MOTA = 88.8%, IDF1 = 91.7%, and HOTA = 65.5%. MCCG with M = 4 retains 95% permissive eligibility while withholding permissive guidance in three of the four ID-switch frames (75%). These findings support component-level feasibility, while leaving field deployment, V2X stack validation, and user studies for future work.

Driving assistance and Autonomous driving are among the fastest evolving domains of intelligent transportation systems (ITSs). However, visually impaired pedestrians (VIPs), remain weakly protected by the roadsides or vehicle perception systems, especially when the right of way must be communicated in an accessible and auditable manner. In this paper we describe a safety-constrained vehicle-to-pedestrian (V2P) architecture designed for VIPs crossing assistance. The system creates a time bounded interaction state between the pedestrians, the roadside infrastructure, and the vehicles at Fog, to avoid permissive instructions before all the safety stacks are respected in each level where there are formal time bounded strictures with appropriate tools to formally abide by these rules during each interaction between vehicles and VIPs. Our objective is not to design a new detector, tracker, or MLLM model. Instead, we combine edge descriptors, fog level multi-view descriptors, and multi-camera multiple objects tracking technique (MC-MOT) within a latency-constrained V2P loop. Our main empirical focus is the WildTrack multi-camera continuity, with a novel Multi-Camera Consensus Gating (MCCG) safety primitive criterion; we also conducted a JAAD experiment as secondary assistive-branch for feasibility check, indicating plausible crossing-intent estimates, sub-second restrictive alerts, and actionable grounded guidance. On WildTrack the reproduced Fog continuity branch reaches a MOTA = 88.8%, IDF1 = 91.7%, and HOTA = 65.5%. In the WildTrack settings with N = 7, MCCG with M = 4 blocks 75% of ID-switch considering a 5.0% permissive-eligibility cost. These findings support feasibility of the proposed ITS interaction loop, while leaving field deployment of V2X stack validation, and user studies for future work.

Unmanned Aerial Vehicle (UAV) networks are increasingly deployed in dynamic environments where reliable and low‐latency communication is critical. However, high mobility, intermittent connectivity, spectrum limitations, and energy constraints make conventional static communication protocols inadequate for maintaining stable, dependable links. To address these challenges, this paper proposes TRC‐MAPPO, a topology‐aware reliability‐constrained multi‐agent deep reinforcement learning framework for adaptive UAV swarm communication. The routing problem is formulated as a constrained decision‐making task that jointly considers packet delivery reliability, end‐to‐end delay, link stability, bandwidth usage, and energy consumption. Unlike single‐agent DRL baselines, TRC‐MAPPO represents the swarm as a dynamic communication graph and combines local relay selection with centralized training, enabling cooperative routing decisions under time‐varying network conditions. Simulations are conducted in a controlled, dynamic UAV environment, with the same mobility and traffic settings used for all methods. The proposed framework is compared with DQN and PPO over different swarm densities. Results show that TRC‐MAPPO achieves a higher packet delivery ratio, lower end‐to‐end delay, and more efficient energy behaviour, particularly in dense deployments. These findings indicate that topology‐aware cooperative learning can provide a scalable and practical solution for reliable UAV communication in future intelligent aerial and 6G‐enabled networks.