Motion Planning for Car-like Robots

Motion planning is one of the most challenging tasks in robotics. Dedicated algorithms are used in many different applications starting from CNC machines to human-like robots. An interesting research area within this field is motion planning for car-like robots. In the last few years, car-like robots became increasingly important, because precise positioning systems like GPS or GLONASS and modern sensor technologies allowed navigating in rural or even in urban terrains. These car-like robots can be used for different kinds of purposes. In the automotive domain, they may pave the way for driving in urban terrain. In the agriculture area, applications for harvesting, fertilizing or lawn mowing can be realized. This thesis focuses on motion planning for car-like robots and particularly on applications where an arbitrary working area should be covered as efficient as possible. The aim of this thesis is to develop a path planning application which is able to compute a feasible path between two arbitrary points. Furthermore, the path shall fully cover a predefined area avoiding obstacles. Moreover, it is desired that the predefined working area is cruised in parallel lanes (e.g., for mowing a soccer field). The path shall be calculated based on lists of geodetic data which represent arbitrary but simple polygons. At the beginning, basic concepts of system theory and system modeling will be studied. Next, different approaches for solving the basic motion planning problem will be discussed. Then, system models for car-like robots will be addressed. Based on such a system model, motion planning concepts are examined which are suitable for car-like robots. Considering the benefits and the drawbacks of all investigated algorithms, a solution based on randomized trees is proposed. Finally, an algorithm for motion planning covering predefined areas in parallel lanes is introduced. A proof-of-concept implementation in combination with a simulation framework allows evaluating the quality and the feasibility of the computed path. The simulation is based on a simple path following controller and a realistic system model for a car-like robot which considers errors of the system (e.g., GPS and heading errors) close to reality. Analyses of the simulation show that the car-like robot is able to follow the computed path even in presence of system errors.