基于长/短视距多模型自适应交互的高速无人车避障规划研究

Under the conditions of low adhesion and high-speed emergency obstacle avoidance, the efficiency of path planning and vehicle dynamics characteristics should be fully considered to ensure that the planned path with the requirements of vehicle high-speed tracking accuracy. To solve the problem of coupling intervention between planning path and tracking control in emergency obstacle avoidance of unmanned vehicles, a local path planning algorithm was proposed based on multi-layer integration of long/short horizon for high-speed emergency obstacle avoidance. Firstly, the risk values of obstacles were refined according to the density of obstacles with different risk values, and the risk potential fields and environmental risk sampling spaces, being capable for cross/circumnavigate obstacle clusters, were designed respectively, so as to construct a long-horizon local path planning algorithm based on bias sampling. Then, a short-horizon local path replanning was proposed based on interactive multi-model predictive control, and the model probability real-time online update rules were established, so as to realize the local path planning of rapid driving under extreme conditions, and meet the requirements of real-time computation, obstacle avoidance safety and tracking stability. Finally, the simulation tests show the effectiveness and superiority of the proposed multi-layer fusion local path planning method.