Multi-Robot Frequency-Based Patrolling

Mobile robots can save human lives and financial costs by replacing humans in mundane, or dangerous tasks. Patrolling is one such task: It is often mundane, is inherently repetitive, and may involve risk to human patrollers. One important type of patrolling is frequency-based patrolling, which involves visiting all points within a target work area at a fixed frequency (as much as possible). Our focus in this dissertation is on frequency-based patrolling by teams of multiple, cooperating, robots. First, it formally defines the frequency-based patrolling problem, and distinguishes three different possible optimization criteria by which to evaluate solutions. Second, it presents solutions that are provably robust to robot death, guaranteeing that patrolling will continue as long as at least one robot is functioning. It also addresses event handling in patrolling, exploring algorithms that optimally select which robot should break the patrolling motion to respond to an event at a given location, and how to distribute the event handling load among different robots. The dissertation makes a clear distinction between patrolling areas (enclosed by polygons), and patrolling polylines. In particular, the dissertation shows that patrolling an open polyline (e.g., a two-ended fence) is inherently in conflict with the goal of achieving uniform point visit frequency. Different algorithms are therefore presented for patrolling areas and for patrolling polylines. Finally, we allow each patrolling unit to include multiple robots that move in formation. Unfortunately, known algorithms for formation maintenance typically utilize the robots’ sensors to carry out the formation, thus prohibiting their use for carrying out the surveillance task. We thus present a first step towards a novel formationmaintenance technique which multiplexes the use of sensors and communications, and allows the robots to utilize their sensors for monitoring their surroundings, while maintaining the formation.