基于行为识别的智能车纵向决策研究

针对智能车纵向决策问题，提出基于环境车辆偏离车道程度识别运动模式的方法；构建动态环境车辆横纵向轨迹预测模型，并求解；构建保持、先行、避让在内的决策集，提出基于预测轨迹的单个车辆决策方法，并基于所有动态环境车辆的决策结果在加速、减速和匀速3 种结果中做出综合决策. 实车实验表明：在直行、换道和转弯运动模式下轨迹预测平均误差分别为0.11，0.29，0.80 m，预测精度较高；复杂动态环境下，本文提供的纵向决策信息提升了智能车行驶的安全性和舒适性.

Intelligent Vehicle Longitudinal Decision Making Based on Behavior Recognition

This paper focuses on the longitudinal decision-making of intelligent vehicles. A method based on the degree of deviation from travel lane was proposed to identify the mode of vehicle in certain environment. Then, the longitudinal and transverse trajectory prediction model was developed considering dynamic traffic environment and being solved. The decision sets include maintaining, leading and avoiding behaviors of vehicle, and the proposed single vehicle decision method was based on the predicted trajectory. Then the comprehensive decision was made for three situations including acceleration, deceleration, and driving with constant speed and considering all dynamic environment. The real vehicle experiment shows that the average error of trajectory prediction in driving straight, lane changing, and making turns is 0.11 meters, 0.29 meters and 0.8 meters, respectively. The results show a high prediction accuracy. The proposed longitudinal decision- making methods can be used to improve the safety and comfort of intelligent vehicle driving under complex dynamic environment.