Title:Adaptive Sampling Control in Motion Planning of Autonomous Vehicle

Abstract:Autonomous driving vehicles aim to free the hands of vehicle operators, helping them to drive easier and faster, meanwhile, improving the safety of driving on the highway or in complex scenarios. Automated driving systems (ADS) are developed and designed in the last several decades to realize fully autonomous driving vehicles (L4 or L5 level).
The scale of sampling space leads to the main computational complexity. Therefore, by adjusting the sampling method, the difficulty to solve the real-time motion planning problem could be incrementally reduced. Usually, the Average Sampling Method is taken in Lattice Planner, and Random Sampling Method is chosen for RRT algorithms. However, both of them don't take into consideration the prior information, and focus the sampling space on areas where the optimal trajectory is previously obtained. Therefore, \emph{in this thesis it is proposed an adaptive sampling method to reduce the computation complexity, and achieve faster solutions while keeping the quality of optimal solution unchanged}.
The main contribution of this thesis is the significant decrease in the complexity of the optimization problem for motion planning, without sacrificing the quality of the final trajectory output, with the implementation of an Adaptive Sampling method based on Artificial Potential Field (ASAPF). In addition, also the quality and the stability of the trajectory is improved due to the appropriate sampling of the appropriate region to be analyzed.