Abstract: Technological advances in the area of unmanned vehicles are opening new possibilities for creating teams of vehicles performing complex missions. However, deploying large mobile sensing networks will require that these vehicles process their collected data and re-plan their missions with as little intervention from human operators as possible.
This talk will describe some approaches to optimizing the performance and trajectories of mobile sensors assuming a dynamic model of the environment. We emphasize the importance of performance bounds, which can drive the search for good heuristics. Scenarios considered include detection, estimation, and data-harvesting missions. Some tools that have proved useful to mitigate the complexity of the computations include relaxations for weakly coupled Markov Decision Processes (bandit problems), and model reduction building on the fluid models used for the control of queuing networks.