Multi-Sensor Resource Management for Layered Sensing

The implementation of ISR architectures such as layered sensing will significantly increase the number and types of sensors available for ATR, but at the same time the ISR problem is becoming become more difficult (e.g., thousands of targets, dismounts, urban terrain, extended operating conditions). To exploit fully these new sensing capabilities, it is necessary to develop multi-sensor resource management algorithms capable of computing near-optimal tasking decisions in real time for thousands of targets and dozens of sensors, and be responsive to changes in sensor performance capabilities versus different target types and operating conditions. To develop such a multi-sensor resource manager we have proposed an innovative combination of generalized multi-armed bandit algorithms and approximate dynamic programming to compute efficient far-sighted solutions of the sensor allocation and scheduling problem. The sensor manager is a model-based approach that can adapt easily to different sensor performance models or optimization criteria, allowing it to respond to different operating conditions and mission requirements. While developing the sensor management algorithm, we will compute bounds on expected optimal ATR fusion performance for the specified level of sensor resources. These performance estimates will be used to quantify the optimality of the algorithms.