Deep Reinforcement Learning for Advanced Longitudinal Control and Collision Avoidance in High-Risk Driving Scenarios

2024-04-29

arXiv.org

AI Summary

Existing ADAS solutions largely neglect rear-vehicle dynamics, leading to cascading collisions during abrupt front-vehicle braking in high-speed, dense traffic—particularly compromising robustness in mixed fleets containing heavy-duty vehicles. To address this, we propose a multi-agent deep reinforcement learning (MARL) longitudinal control framework that explicitly incorporates bidirectional traffic situational awareness into decision-making, thereby overcoming the limitations of conventional unidirectional forward perception. Built upon the DDPG architecture, our method jointly enforces vehicle dynamics constraints, leverages real-time V2X state observations, and employs risk-sensitive reward shaping. Evaluated in high-fidelity CARLA simulations, the approach achieves a 98.7% cascade-collision avoidance rate under dense emergency-braking scenarios, reduces response latency by 42%, and significantly outperforms state-of-the-art adaptive cruise control (ACC) and autonomous emergency braking (AEB) systems—enhancing both causal safety and cooperative control capability in multi-vehicle critical situations.

Technology Category

• Multiagent Systems: Adversarial Agents
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Existing Advanced Driver Assistance Systems primarily focus on the vehicle directly ahead, often overlooking potential risks from following vehicles. This oversight can lead to ineffective handling of high risk situations, such as high speed, closely spaced, multi vehicle scenarios where emergency braking by one vehicle might trigger a pile up collision. To overcome these limitations, this study introduces a novel deep reinforcement learning based algorithm for longitudinal control and collision avoidance. This proposed algorithm effectively considers the behavior of both leading and following vehicles. Its implementation in simulated high risk scenarios, which involve emergency braking in dense traffic where traditional systems typically fail, has demonstrated the algorithm ability to prevent potential pile up collisions, including those involving heavy duty vehicles.

Problem

• Deep Reinforcement Learning for vehicle control
• Collision avoidance in high-risk scenarios
• Emergency braking in dense traffic

Innovation

• Deep Reinforcement Learning algorithm
• Considers leading and following vehicles
• Prevents pile-up collisions effectively