Path planning for an autonomous racecar with Rapidly-exploring Random Tree

Among the considerable technical problems that self-driving vehicles have to face, there is the path planning, that is a function for autonomous vehicles. Path planning is defined as the problem of finding a continuous collision free path from an initial configuration to a predetermined goal. In this context, the present thesis work focuses on the design and implementation of a real-time local trajectory planning method using Rapidly-Exploring Random Tree Algorithm based on Dubins curves for an AWD electric racing vehicle in different scenarios, straight, left and right bends. The key contribution of this work is to present a framework able to generate a feasible free collision path considering differential constraints for non-holonomic car-like robot and its real-time computation ability in solving the dynamic motion planning problem. As a first step, a preliminary research on motion planning techniques is carried out. Subsequently, the problem is stated, analysing the structured environment in which the race car has freedom of action. Afterwards, the methodology is defined and the final configuration, i.e. the goal, is extrapolated from the local map coming from the perception pipeline, exploiting the functionality of a LiDAR sensor mounted onto the front wing of the racing vehicle. The vehicle is considered as a three Degree-of-Freedom bicycle dynamic model and a Stanley Controller is implemented to control the lateral and longitudinal vehicle dynamics. For validation purposes, the motion planning model has been simulated by means of Mathworks SIMULINK software environment and afterwards implemented on dSpace MicroAutobox. The feasibility of the planned trajectory is also evaluated with respect to command signals for the steering and acceleration actuators featured by the retained racing vehicle. Conclusions with critical comments about the obtained results and the possible future works perspectives complete the present thesis work.