Swarming algorithms for multi-robot systems

A large effort has been devoted by the scientific community to the field of multi-robot systems. The main reason relies on the fact that they exhibit better fault-tolerance, flexibility and performance than a single robot unit. In this thesis, novel contributions to this field are given. Novel decentralized swarming algorithms integrated with obstacle avoidance techniques where the interaction is assumed to be limited by a range of view are presented. Furthermore, to better comply with the hardware/software limitations of mobile robotic platforms, the actuators of the robots are assumed to be saturated. Due to technological advances in the field of electronic devices, the robots are able to share data in a fast and reliable manner among themselves. This capability is used within this thesis to carry out a distributed estimation of system wide quantities to adjust the interactions of each individual with its neighbors. In particular, the average consensus problem over digraphs is addressed and a distributed strategy to solve it is presented. Moreover, a finite time condition to check if the communication digraph is strongly connected or it is not is detailed. It can be verified in a distributed fashion by the robots and can be employed to avoid useless steps in the estimation process reviewed within this work. A theoretical characterization of the properties of the presented approaches is provided. Moreover, experiments in real scenarios using a team of low-cost mobile robots SAETTA (built in our laboratory) to demonstrate the effectiveness of the proposed algorithms have been carried out. Finally, a relative distance and a relative localization system are introduced to provide the indispensable input for the swarming algorithms