VS-DDP: Variable Splitting Based Differential Dynamic Programming Algorithm for Autonomous Trajectory Planning

In this article, we introduce a novel trajectory planning approach for autonomous vehicle, called as Variable Splitting based Differential Dynamic Programming (VS-DDP). Utilizing variable splitting method, VS-DDP can successfully handle the constraints presented in trajectory planning problem, which extremely extends the power of differential dynamic programming. Simulation and real world experiments are conducted to validate the effectiveness and efficiency of VS-DDP. With a comparison between VS-DDP and direct collocation based method existing in famous Apollo project [1], we find that VS-DDP greatly outperforms the comparison method in terms of the computation efficiency.