Cooperative Autonomous Transportation Systems: Embodied Multi-Agent Intelligence from Simulation to Real-World Deployment

Transportation is moving from independently operated vehicles toward connected, mixed-autonomy systems involving autonomous cars, buses, trucks, delivery robots, roadside units, traffic signals, and human drivers. These participants continuously affect one another, so safe and efficient mobility requires closed-loop intelligence that links perception, communication, decision-making, and physical control. Embodied intelligence provides this interaction-centered perspective, while multi-agent learning offers tools for coordinating heterogeneous agents with shared or competing objectives. Recent progress in foundation models, world models, reinforcement learning, vehicle-to-everything communication, digital twins, and high-fidelity simulation has expanded the possibilities for cooperative transportation research. However, many methods remain limited to single vehicles, homogeneous agents, simplified scenarios, or laboratory-only evaluation. A transportation-focused framework is needed to connect algorithmic advances with realistic traffic environments, reproducible benchmarks, safety assurance, and reliable transfer to road, logistics, and public-transport applications.This Research Topic aims to address the gap between isolated autonomous-vehicle algorithms and coordinated transportation systems in which vehicles, roadside infrastructure, mobile robots, and human-driven traffic interact. Existing methods often optimize perception, planning, traffic control, or reinforcement learning for a single agent or a fixed simulation, making them difficult to scale to heterogeneous traffic participants and to validate safely before deployment. We seek integrated approaches in which embodied agents perceive road scenes, communicate with other agents and infrastructure, reason over long horizons, and jointly plan and control their actions. Recent advances in multi-agent reinforcement learning, foundation and world models, vehicle-to-everything cooperation, digital twins, and high-fidelity simulators provide new opportunities to build adaptive transportation intelligence. The Topic will encourage common benchmarks, realistic traffic scenarios, interoperable simulation platforms, robust sim-to-real transfer, and rigorous evaluation of safety, scalability, uncertainty, explainability, and human interaction. Its goal is to connect transportation algorithms with reusable platforms and real-world demonstrations.This Research Topic focuses on cooperative autonomous transportation and welcomes original research articles, reviews, perspectives, methods papers, benchmark studies, and application-oriented case studies. Themes include multi-agent perception, communication, planning, reinforcement learning, and distributed control for vehicles and traffic networks; embodied foundation models and world models for driving and mobility; vehicle-to-vehicle and vehicle-to-infrastructure cooperation; coordination among cars, buses, trucks, delivery robots, drones, roadside units, and human drivers; traffic-signal and network control; digital twins, simulation and testing platforms; scenario generation, datasets, and benchmarks; sim-to-real transfer and edge-cloud collaboration; and safety, robustness, interpretability, privacy, cybersecurity, and human-agent interaction. Submissions should define the transportation problem and embodied agents clearly, explain their coordination mechanisms, and provide rigorous evaluation through realistic simulation, reproducible benchmarks, field experiments, or real-world deployment. Work connecting algorithms to reusable platforms is particularly encouraged. Topic editors: Nuo Lei, Zirui Li, Hao E. Zhang Source: https://www.frontiersin.org/research-topics/86669/cooperative-autonomous-transportation-systems-embodied-multi-agent-intelligence-from-simulation-to-real-world-deployment