基于客观风险感知特性的前向碰撞预警算法优化与标定

为提升高级驾驶辅助系统(ADAS)预警算法在复杂行车环境下的适应性, 提出了一种基于车辆运动学和风险感知特性的综合预警算法——客观风险感知(ORP)算法; 通过典型工况下的分析与推导, 表明其预警算法为THW、TTC和SM预警算法的综合形式; 为了标定预警算法的参数阈值, 开展了累计4 500 km自然驾驶试验, 最终筛选出409例有效临近碰撞事件, 提取了释放油门、踩下制动时刻的客观风险感知参数分布特征; 根据自然驾驶数据中提取的临近碰撞事件及其参数特征, 对风险预警算法参数进行标定; 在模拟驾驶环境下开发了前向碰撞预警算法, 通过4种风险场景开展了算法验证试验。研究结果表明: 基于自然驾驶数据的参数标定, 客观风险感知预警算法的两级预警参数阈值分别为1.4、0.8 s; 基于典型风险工况下的微观驾驶行为特性对比, 预警有效性方面ORP预警算法稍高于RP预警算法, 二者预警有效性显著高于TTC预警算法; 在预警算法下所有驾驶片段的最小碰撞时间均值方面, ORP预警算法为2.02 s, RP预警算法为1.90 s, TTC预警算法为1.65 s, 表明ORP预警算法能适应复杂风险环境下的风险辨识。基于大量实车试验参数标定与效果验证后, 所提出预警算法可用于高级驾驶辅助系统风险辨识。 

Forward collision warning algorithm optimization and calibration based on objective risk perception characteristic

In order to improve the adaptability of advanced driver assistance system(ADAS) warning algorithm in complex driving environments, a comprehensive warning algorithm named the objective risk perception(ORP) algorithm based on the vehicle kinematics and risk perception characteristic was proposed. The analysis and derivation under typical risk conditions show that the proposed warning algorithm is a comprehensive mode of time headway(THW), time-to-collision(TTC) and safety margine(SM) based warning algorithms. In order to calibrate the parameter thresholds of the proposed warning algorithm, a total of 4 500 km natural driving experiments were carried out, and finally 409 valid near-crash events were extracted. The distribution characteristics of objective risk perception parameters when release accelerator and press brake were obtained. The risk warning algorithm parameters were calibrated based on the near-crash events and their parameter characteristics extracted from the natural driving data. The forward collision warning algorithm was developed under a simulated driving environment, and the verification experiments of the algorithm were carried out based on four risk scenarios. Research result shows that based on the parameter calibration of natural driving data, the two-level warning parameter thresholds of the ORP warning algorithm are 1.4 and 0.8 s, respectively. Based on the comparison of driving behavior under typical risk conditions, in terms of warning effectiveness, the ORP warning algorithm is slightly higher than the RP warning algorithm, and the effectiveness of secondary early warnings is significantly higher than that of the TTC warning algorithm. In terms of the average minimum time-to-collision of all driving segments under the warning algorithm, the ORP warning algorithm is 2.02 s, the RP warning algorithm is 1.90 s, and the TTC warning algorithm is 1.65 s, which shows that the ORP warning algorithm can adapt to the risk identification in complex risk environment. Based on a large number of natural driving test based parameter calibration and effect verification, the proposed warning algorithm can be used for the risk identification of advanced driver assistance systems.