基于安全风险预测的自动驾驶自适应巡航控制优化

Adaptive Cruise Control Optimization of Automatic Driving Based on Safety Risk Prediction

摘要:

从周边车辆运动学状态参数和道路设施条件参数中提取场景特征指标和安全风险度量指标，采用极端梯度提升模型（XGboost）和长短时记忆模型（LSTM）进行安全风险预测，由此提出基于安全风险预测的自动驾驶自适应巡航控 制（ACC）优化方法，并选取碰撞发生概率、速度平均值、速度标准差3种指标评价ACC表现。通过Prescan和Simulink联合仿真推演，验证了ACC优化方法的合理性和有效性。结果表明，基于安全风险预测的ACC优化方法的控制表现优于一般ACC；利用LSTM预测安全风险，相比XGboost具有更好的ACC优化表现；预测安全风险时增加道路设施条件参数，显著提升了ACC表现，降低了自动驾驶碰撞发生概率。

Abstract:

This paper， extracting the scenario feature indexes and risk metrics index from vehicle kinematic status parameters and road infrastructure condition parameters， uses the extreme gradient boosting （XGboost） model and the long short-term memory （LSTM） model for safety risk prediction. Then， it proposes an adaptive cruise control （ACC） optimization method of automatic driving based on safety risk prediction. It selects collision probability， average speed， and standard deviation of speed to evaluate the performance of ACC optimization， and verifies the rationality and effectiveness of the ACC optimization method proposed using Prescan-Simulink co-simulation. The results show that the safety risk-based ACC optimization method is superior to the general ACC. Compared with the XGboost， the LSTM as safety risk prediction model， has a better performance for ACC optimization. The addition of road infrastructure condition parameters for safety risk prediction improves the ACC performance and reduces the collision probability of automatic driving significantly.