DynaPA: Dynamic Power Allocation for Improved Exploration-Exploitation in Active Sensing

This paper addresses an adaptive active sensing problem of nonstationary sparse multi-target detection, by dynamically shaping the transmit beampattern of a MIMO radar. In order to strike a better balance between exploration and exploitation (two key ideas in any online decision making problem), a dynamic power allocation algorithm called DynaPA is developed that can redirect exploration energy for exploitation (and vice versa) depending on the environment. Numerical results demonstrate superior performance of DynaPA over existing algorithms with varying target numbers.