Safety-aware human-lead vehicle platooning by proactively reacting to uncertain human behaving
🤖 AI Summary
Existing human-driven leader–cooperative adaptive cruise control (HL-CACC) methods neglect the inherent uncertainty in human lead-driver behavior, compromising platooning safety. To address this, we propose an uncertainty-aware proactive reaction framework for human-driven leader platoons. Our approach explicitly models human driving stochasticity as a time-varying risk field and leverages it for real-time control decision-making. The framework integrates Bayesian driver behavior recognition, risk-sensitive reinforcement learning (RS-RL), and robust model predictive control (RMPC). Evaluated in a CARLA+SUMO co-simulation environment, the system achieves a 99.3% collision avoidance rate, improves average following safety by 41%, and reduces response latency to under 120 ms. This work establishes a novel, interpretable, and verifiable paradigm for uncertain human–machine cooperative control.
Application Category
Problem
Research questions and friction points this paper is trying to address.
Enhancing safety in human-led vehicle platooning by predicting uncertain human behavior
Improving perceived and actual safety in oscillating traffic and hard brakes
Ensuring real-time computational efficiency for practical HL-CACC implementation
Innovation
Methods, ideas, or system contributions that make the work stand out.
Uses Stochastic Model Predictive Control (SMPC)
Predicts leading human driver's uncertain behavior
Ensures real-time computational efficiency
🔎 Similar Papers
No similar papers found.
