Stephen L. Smith, Department of Electrical and Computer Engineering, University of Waterloo
Optimizing Motion for Robotic Monitoring and Information Gathering
Robots are increasingly being used in long-term monitoring tasks such as environmental monitoring, infrastructure inspection, and information gathering. However, to enable operation in large-scale environments there is a need for provably efficient methods for planning robot paths. In discrete environments, the path planning problems are combinatorial in nature and look similar to the well known vehicle routing problem. Unfortunately, the optimization objectives which capture monitoring tasks have not been considered in the vehicle routing literature. Thus, this talk will consider two new optimization objectives for monitoring tasks in discretized environments; the min-max latency walk problem, which captures the monitoring of changing environments, and the submodular traveling salesman problem problem, which captures information gathering tasks. For both objectives, the computation of an exact solution is intractable and thus approximation algorithms will be developed. The algorithms make interesting connections to known combinatorial and submodular optimization problems, and can be applied to large-scale environments.