An Uncertainty-Aware Lane Change Motion Planning Algorithm Based on Probabilistic Trajectory Prediction Distribution

Comprehensive and accurate understanding of the interactive traffic environment facilitates reasonable motion planning for automated vehicles. This paper presents an overall risk assessment method for the host vehicle to achieve efficient motion planning considering uncertainties of the predicted driving behaviors of surrounding vehicles. A Social Temporal Convolutional Long Short-Term Memory network is constructed to capture the interactive characteristics among the host and surrounding vehicles and to predict the statistical distribution of the trajectory prediction uncertainty in the prediction horizon. Then a two-dimensional Gaussian distribution-based dynamic risk assessment with a soft update method is developed to spatially and temporally quantify the driving risk by constructing the occupancy map based on the multi-modal distribution of the predicted trajectories for the surrounding vehicles. The optimal motion of the host vehicle is determined by minimizing a multi-objective function of the alternative driving behaviors. The effectiveness of the proposed scheme is verified under typical driving scenarios extracted from the NGSIM dataset. The results show that the proposed method can comprehensively evaluate the potential risk and efficiently achieve motion planning while minimizing the driving risk.