Control of Mobile Robots Moving in Cluttered Environments
Over the past few decades, mobile robots have gained a lot of attention, particularly with the evolution of application fields such as search and rescue, cleaning, and exploration. Developing such robots requires to cope with different challenges such as perception, tracking, and mapping. Nevertheless, regardless of the mission to be performed or the application domain, robots must be able to plan their own motion. Hence, motion planning is at the heart of robotics and has been thoroughly addressed since the first mobile robot was developed. Usually, real-world environments are unknown and change over time. Therefore, traditional path planning methods that build upon a previously known map fail to work properly in these environments. Reactive collision avoidance approaches tackle this problem by incorporating the perceived information into the control system, bridging the gap between planning a path and executing a motion. Unfortunately, the majority of these methods undergo some classical drawbacks limiting their performance in cluttered environments. These include being prone to oscillations, failure of guiding a robot through narrow spaces, neglect of the robot constraints, and the tendency to generate longer paths and higher execution times. The work presented in this thesis aims to cope with the above mentioned problems. To this end, a novel collision avoidance approach was developed and implemented. The key idea is to analyze the environmental structure and find out the most promising gap, once determined, a subgoal is located in a collision-free area. It is located in such a way that the opening angle of the selected gap is considered, providing a safer and smoother bridge between collision avoidance and target approach. This also leads to shorter paths and less
execution times. The proposed approach has been improved by considering the clearance to obstacles and by computing the steering angle in such a way that all surrounding obstacles are taken into account. This has been possible by introducing and integrating two concepts, called “tangential” and “gap flow” navigation. Another contribution is the computation of the motion command in such a way that the stability of the system is guaranteed in the Lyapunov sense. Furthermore, this work presents a new concept, the “admissible gap”, which addresses the question of whether a given gap is traversable by performing an admissible collision-free motion control. This concept has been successfully employed to develop a collision avoidance approach, that directly respects the vehicle constraints rather than adapting a holonomic-based solution. Another contribution is the development of a new strategy for extracting gaps, which reduces the possibility of oscillation and improves the stability of navigation. Finally, experimental results along with performance assessment in highly cluttered scenarios are presented to verify that the proposed approaches outperform state-of-the-art techniques in terms of smoothness, efficiency, reliability, and safety.