Locating RF Emitters with Large strongUAVstrong Teams.pdfVIP

Locating RF Emitters with Large UAV Teams Paul Scerri, Robin Glinton, Sean Owens and Katia Sycara School of Computer Science Carnegie Mellon University Pittsburgh, PA 15213, USA {pscerri, rglinton, owens, katia}@cs.cmu.edu April 24, 2007 Abstract This chapter describes a principled, yet computationally efficient way for a team of UAVs with Received Signal Strength Indicator (RSSI) sen- sors to locate radio frequency emitting ground vehicles in a large envi- ronment. Such a capability has a range of both civilian and military applications. RSSI sensor readings are noisy and multiple emitters will cause ambiguous, overlapping signals to be received by the sensor. Gener- ating a probability distribution over emitter locations requires integrating multiple signals from different UAVs into a Bayesian filter, hence requir- ing cooperation between the UAVs. To build a coherent distributed pic- ture given communication limitations, the UAVs share only those sensor readings that induce the largest changes in their local filter. Each UAV translates its probability distribution into a map of information entropy and then plans a path that will maximize the reduction in entropy (or con- versely provides the highest information gain.) Planned paths are shared with a subset of other UAVs to minimize overlapping search. Experi- ments in a medium fidelity simulation environment show the approach to be lightweight and effective. Live flight results with lightweight Class I UAVs validate our approach. 1 Introduction The rapidly improving availability of small, unmann