Control of Outdoor Robots at Higher Speeds on Challenging Terrain

This thesis studies the motion control of wheeled mobile robots. Its focus is set on high speed control on challenging terrain. Additionally, it deals with the general problem of path following, as well as path planning and obstacle avoidance in difficult conditions. First, it proposes a heuristic longitudinal control for any wheeled mobile robot, and evaluates it on different kinematic configurations and in different conditions, including laboratory experiments and participation in a robotic competition. Being the focus of the thesis, high speed control on uneven terrain is thoroughly studied, and a novel control law is proposed, based on a new model representation of skid-steered vehicles, and comprising of nonlinear lateral and longitudinal control. The lateral control part is based on the Lyapunov theory, and the convergence of the vehicle to the geometric reference path is proven. The longitudinal control is designed for high speeds, taking actuator saturation and the vehicle properties into account. The complete solution is experimentally tested on two different vehicles on several different terrain types, reaching the speeds of around 6 m/s, and compared against two state-of-the-art algorithms. Furthermore, a novel path planning and obstacle avoidance system is proposed, together with an extension of the proposed high speed control, which builds up a navigation system capable of autonomous outdoor person following. This system is experimentally compared against two classical obstacle avoidance methods, and evaluated by following a human jogger in outdoor environments, with both static and dynamic obstacles. All the proposed methods, together with various different state-of-the-art control approaches, are unified into one framework. The proposed framework can be used to control any wheeled mobile robot, both indoors and outdoors, at low or high speeds, avoiding all the obstacles on the way. The entire work is released as open-source software.