Models and algorithmic approaches for cooperative multi-robot systems: thesis abstract

The subject of multi-robot systems has been thoroughly investigated in the past decade (e.g., [9, 12, 11, 7]). Current research in multi-robot systems can be divided, roughly, according to two aspects: theoretical robotics (e.g. [12]) and practical robotics (e.g. [10]). Theoretical robotics research, in many cases, make unrealistic assumptions on the robots’ capabilities. For example, robots are sometime assumed to have unlimited visibility [8], or limited radius of visibility but omnidirectional [6]. In practice, it is usually not the case — robots have a limited angle of visibility along with limited radius of visibility. On the other hand, practical robotics research is driven, in many cases, by the will to solve problems, thus tend to be proven only for a specific robotic system. In my work, similar to work done by several other researchers, I have been aiming to bridge the gap between these two disciplines by examining problems from the multirobot domain using theoretical tools, i.e., trying to model the environment and provide computational analysis of problems and solutions while taking under consideration sensorial, motion or other limitations of the robots. My work includes three topics. The first, multi-robot task allocation based on the cost of interaction between the team members. In the second part of my work I handled the multirobot coverage task in which the goal was to find coverage paths such that the total time of coverage is minimized. The third topic is derived from the coverage task, and concerns multi-robot patrolling (repeated coverage) in adversarial environments. All problems are dealt using theoretical tools, for example using graph representation, graph theory algorithms, etc. The algorithms are evaluated using complexity measures, and optimality of the algorithms is theoretically proven (for non-heuristic solutions). The following sections describe the work done in each topic in more details.