车车通信环境下基于驾驶意图共享的车辆避撞预警算法

针对传统基于距离或时间的车辆避撞预警算法存在较高误警率的问题,考虑在避撞预警算法中引入驾驶意图共享的概念,提出了基于实际外场复杂车车通信V2V（Vehicle-to-vehicle,V2V）环境下的车辆跟驰避撞预警算法。基于LTE-V构建外场V2V环境,将车辆行驶过程描述为一个时间序列的隐性马尔可夫随机过程,借助隐马尔可夫模型（Hidden Markov Model,HMM）建立驾驶人驾驶意图与车辆相对行驶状态序列之间的隐含关系模型,并给出基于Viterbi算法的驾驶意图预判求解方法,将驾驶意图作为特征因子集成到安全距离模型中,提出基于驾驶意图共享的避撞（Driving Intention Based Collision Avoidance,DI-CA）预警算法。利用构建的V2V试验环境,实现了匀速、加速、减速和紧急制动等4种驾驶意图,以及相对速度和相对距离增加、减小、保持不变等9种组合的车辆行驶状态试验数据获取,并利用试验数据对所提出的DI-CA预警算法进行实证分析。结果表明：所提出预警算法能够针对不同驾驶意图提供有效的车辆碰撞预警。在此基础上将4种驾驶意图下的DI-CA预警算法与Mazda预警算法求得的安全距离进行了对比分析,所提出的DI-CA预警算法的平均预警正确率为84%,高于Mazda预警算法的78%,而DI-CA预警算法的平均误警率和漏警率分别为5%和16%,均明显低于Mazda预警算法,说明所提出的DI-CA预警算法在提升预警效果的同时明显降低了误警率和漏警率,可避免行驶过程中因误警而导致的连续刹车,以及因漏警而导致的可能碰撞事故发生。最后,总结并给出了驾驶意图共享理论应用于车辆避撞预警的研究展望。

Warning Algorithm of Vehicle Collision Avoidance Based on Driving Intention Sharing in Vehicle-to-vehicle Environment

Aiming at solving the problem of high false alarm rates of traditional early-warning algorithms of vehicle collision avoidance based on distance or time, the concept of driving intent sharing was considered in this study, and a collision avoidance early-warning algorithm based on a complex vehicle-to-vehicle (V2V) environment was proposed. Based on a long term evolution-vehicle(LTE-V) technology, an outfield V2V environment was constructed, and the vehicle driving process was described as a time-series hidden Markov random process. By using a hidden Markov model (HMM), the implicit relationship between a driver's driving intention and the relative driving state sequence of the vehicle was established. The method for predicting the driving intention based on a Viterbi algorithm was also presented. The driving intent was integrated into the safety distance model as a characteristic factor, and a driving intention-based collision avoidance (DI-CA) warning algorithm was proposed. By considering the built V2V experimental environment, four driving intentions of "constant speed," "acceleration," "deceleration," and "emergency braking" were achieved, and the relative speed and relative distance were "increased," "decreased," or "maintained" to obtain experimental data of nine combinations of vehicle driving conditions, such as "change." An empirical analysis of the proposed DI-CA early-warning algorithm based on the experimental data shows that the proposed early-warning algorithm can provide effective vehicle collision early-warning for different driving intentions. Based on the analysis, the safety distances obtained using the DI-CA early warning algorithm and the Mazda early-warning algorithm for four driving intentions were compared and analyzed. The average early-warning accuracy rate of the proposed DI-CA early-warning algorithm is 84%, higher than that of the Mazda early-warning algorithm (78%). The average false alarm rate and missed alarm rate of the DI-CA early-warning algorithm are 5% and 16%, respectively, which are significantly lower than those of the Mazda early-warning algorithm. The results show that the proposed DI-CA early-warning algorithm improves the early-warning effect and significantly reduces the false alarm rate and missed alarm rate, which can minimize the continuous braking caused by false alarms during driving and possible collision accidents caused by missed alarms. Finally, the research prospects of the driving intention sharing theory applied to vehicle collision avoidance warning are highlighted.