Optimum Power Allocation for Sensor Networks That Perform Object Classification

In this publication, the power allocation problem for a distributed sensor network is formulated as a signomial program, and analytically solved by a Lagrangian setup. Typical examples of such networks are active radar systems with multiple nodes whose aim is to detect and classify target objects. As it is common for sensors with weak power-supplies, constraints by sum and individual power limitations are imposed. For each sensor node, an amplify-and-forward strategy for the reflected and received echo is proposed. This per-node information is transmitted over a communication channel and combined at a fusion center. The fusion center carries out the final decision about the type of the target object by a best linear unbiased estimator and a subsequent distance classification. In contrast to approaches in the literature, which combine discrete local decisions into a single global one, the approach in the current paper offers many advantages, ranging from the simplicity of its implementation to the achievement of an optimal solution in closed-form and design of the sensor network.