基于反应式行为的车辆运动意图辨识

Identification of vehicle motion intention based on reaction behavior model

为提高无人驾驶智能汽车决策规划能力,提出了一种基于反应式行为的车辆运动意图辨识策略。对反应式车辆运动意图进行了分析与归类,建立了基于多维高斯隐马尔科夫过程的反应式车辆运动意图模型;搭建了车辆运动意图信息采集系统,进行了典型工况数据采集;在此基础上,对模型进行训练,利用正交试验对模型参数进行优化,选取典型工况进行辨识测试。结果表明,所提出的辨识策略能够有效辨识交通车辆的运动意图。

To improve the performance of decision-making planning for intelligent vehicles, a vehicle motion intention identification strategy is proposed based on the reactive behavior. Analysis and classification are performed for reactive vehicle motion intention, and a Reaction Motion Intention Model （RMIM） is built based on Multi-dimension Gaussian Hidden Markov Process （MGHMP）. The experimental data of typical cases are obtained from the vehicle motion intention data acquisition system, and RMIM is trained and identified. The key parameters of the model are optimized using orthogonal test method, and typical cases are selected for measurement and analysis. Test results show that the reaction intention of traffic vehicle can be identified effectively using the proposed identification strategy.