Autonomous mobile robot systems are gaining more and more importance in a wide variety of applications. The fundamental tasks of these systems include the capability of navigating from a specific position to a desired goal position.
The topic of this BSc Thesis is the control of autonomous car-like vehicles. More specifically the algorithms which are capable of computing continuous curvature trajectories for these vehicles in a known workspace. The capability of following these paths makes the robots more efficient regarding its movement. The workspace may contain any number of obstacles of any size. Therefore, the algorithm has to recognize if a collision-free trajectory does not exist, or else, it has to find a collision-free path.
First, I explain the fundamentals of controlling a car-like vehicle. Next, I briefly explain the different approaches of this path planning problem through different algorithms. Then, I expound the Rapidly-exploring Random Tree algorithm. Utilizing this, I implement an application in MATLAB environment, which is capable of computing a collision-free continuous curvature trajectory for the defined robot model. Then, I explain the Bidirectional Rapidly-exploring Random Tree algorithm, which is an improved version of the previous one. Utilizing this, I implement a new continuous curvature path planning application in MATLAB environment.
To conclude I show results of the two different implementations for several workspaces. I compare the results of the simulations, and highlight the differences in performance.