Biased-Attention Guided Risk Prediction for Safe Decision-Making at Unsignalized Intersections

TL;DR

This paper tackles proactive safety and efficiency at unsignalized intersections by integrating a Transformer-based risk predictor with a Soft Actor-Critic (SAC) DRL framework. The risk predictor uses biased attention to produce a dense risk reward that guides the agent toward safer, faster crossing decisions, while a hierarchical experience replay buffer accelerates learning. Empirical results in a SUMO-based simulation show significant improvements in Average Waiting Time, Average Queue Length, and Collision Rate over rule-based and other RL baselines. The combination of biased risk perception, dense rewards, and structured replay demonstrates a robust approach to proactive safety in complex traffic scenarios.

Abstract

Autonomous driving decision-making at unsignalized intersections is highly challenging due to complex dynamic interactions and high conflict risks. To achieve proactive safety control, this paper proposes a deep reinforcement learning (DRL) decision-making framework integrated with a biased attention mechanism. The framework is built upon the Soft Actor-Critic (SAC) algorithm. Its core innovation lies in the use of biased attention to construct a traffic risk predictor. This predictor assesses the long-term risk of collision for a vehicle entering the intersection and transforms this risk into a dense reward signal to guide the SAC agent in making safe and efficient driving decisions. Finally, the simulation results demonstrate that the proposed method effectively improves both traffic efficiency and vehicle safety at the intersection, thereby proving the effectiveness of the intelligent decision-making framework in complex scenarios. The code of our work is available at https://github.com/hank111525/SAC-RWB.

Figures (10)

• Figure 1: Schematic Diagram of the Intersection Scenario
• Figure 2: Schematic of the decision framework. Note: OOC stands for Out of Control (vehicle has lost control).
• Figure 3: Biased Attention Structure
• Figure 4: Biased Attention Visualization. (a) Raw Attention Scores ($QK^T/\sqrt{d_k}$). (b) Scores with Bias ($QK^T/\sqrt{d_k} + B$). (c) Attention Bias Matrix.
• Figure 5: Convergence Analysis of the Model Training Process (a) Conflict Rate Convergence. (b) Average Reward Convergence. Note: Vehicles with speed less than 0.1 m/s are considered to have collided.