Multi-lane Lane Change Decision and Trajectory Planning of Vehicle Based on Logic Regression Algorithm and Gaussian Probability Density Model 2022-01-7070

Aiming at the trajectory planning of autonomous vehicles in lane changing under muti-lane traffic scenarios, a multi-lane lane change decision and trajectory planning algorithm based on Logic Regression Algorithm and Gaussian Probability Model is proposed. Firstly, the target state (time and velocity) of vehicle during lane change is sampled and lateral trajectory (5^th polynomial) and longitudinal trajectory (4^th polynomial) are planned based on the target state; Secondly, cost evaluation function of trajectory is established and optimal trajectory is selected based on three aspects of safety, comfort and lane changing time. In the assessment of safety, response weighting function is established to characterize the ability of autonomous vehicle to respond to accidents along the path, and the comfort is evaluated by the lateral and longitudinal jerk; Farther, Gaussian probability density model is used to predict the state of target vehicle (TV) in the target lane, and in order to improve the lane change safety, the intention of vehicle in the side lane of target lane(STL) entering the target lane is estimated based on the logistic regression function; Finality, Simulink & Prescan simulations were conducted to verify this algorithm performance. The results indicate that according to the method proposed in this paper, autonomous vehicles can conduct a reasonable risk assessment of the current lane change scenario, comprehensively consider the lane change safety of the target lane vehicle and the cut-in of the side lane of target lane vehicle, make a safe decision-making behavior, and obtain the comprehensive optimal lane change trajectory.