Adaptive spectrum sensing and estimation

We propose a multistage adaptive approach to spectrum sensing and estimation with the goal of concentrating more sensing resources on spectral components of interest. The allocation of resources to minimize the mean squared estimation error is formulated as a dynamic program. An optimal policy is given for the case of two sensing stages. For more than two stages, tractable approximate policies are developed based on open-loop feedback control (OLFC). These policies improve monotonically with the number of stages, and in particular upon the optimal two-stage policy. A spectrum sensing simulation shows substantial reductions in mean squared error compared to non-adaptive sensing and a recently proposed adaptive method. Performance gains in detecting unoccupied channels are also shown.