Integrating Machine Learning for Enhanced Autonomous Driving and V2I Communication

The recent advancements in the field of autonomous driving and vehicular communication have significantly shifted towards integrating advanced machine learning models with traditional reinforcement learning techniques for optimizing both vehicle-to-infrastructure (V2I) communication and autonomous driving policies. This integration aims to enhance traffic flow, improve road safety, and reduce communication handovers. Notably, the use of large language models (LLMs) for autonomous driving decision-making, combined with double deep Q-learning algorithms (DDQN) for V2I optimization, has shown promising results in simulations, demonstrating faster convergence and higher rewards compared to conventional methods. Additionally, the importance of infrastructure in collaborative perception has been re-evaluated, with studies showing that incorporating infrastructure data can significantly improve 3D detection accuracy and enhance noise robustness. The field is also witnessing a move towards dynamic benchmarks for performance evaluation, which adjust for spatial and temporal variations, providing more equitable comparisons between automated driving systems and human-driven fleets. Furthermore, the concept of augmented intelligence through local digital twins at smart intersections is emerging as a key strategy to improve safety and efficiency in autonomous driving by leveraging roadside units for real-time cooperative perception and local agent support. Lastly, the challenge of transmitting large data objects wirelessly under safety constraints is being addressed through innovative application-aware mechanisms, ensuring reliable and low-latency data exchange in autonomous mobile systems.

Sources

Hybrid LLM-DDQN based Joint Optimization of V2I Communication and Autonomous Driving

Dynamic Benchmarks: Spatial and Temporal Alignment for ADS Performance Evaluation

V2I-Calib++: A Multi-terminal Spatial Calibration Approach in Urban Intersections for Collaborative Perception

Rethinking the Role of Infrastructure in Collaborative Perception

Augmented Intelligence in Smart Intersections: Local Digital Twins-Assisted Hybrid Autonomous Driving

Characterizing Behavioral Differences and Adaptations of Automated Vehicles and Human Drivers at Unsignalized Intersections: Insights from Waymo and Lyft Open Datasets

Wireless Large Object Transmission Under Safety Constraints (LOTUS)

Built with on top of