Abstract
This paper focuses on the design of an efficient hierarchical control system for a self-driving car in order to navigate safely in the urban environment. The proposed technique is composed of motion planning, local path planner, and control. Firstly, the decision making is designed by applying two-stages finite state machine (FSM) which manipulates the mission planning and control states on the road. Then, the local path planner is conducted to generate a safe and comfortable trajectory. Moreover, an optimization problem with nonlinear constraints is defined and solved to optimize the jerk. Besides, we perform a real-time hybrid A* algorithm based on a probabilistic occupancy grid map to avoid obstacles. Finally, controllers are designed by using the adaptive-purse pursuit controller for lateral control and the scheduled feed-forward PID controller for longitudinal direction. The experimental results show that our algorithms can work efficiently in a practical scenario.