Optimization of Binary Integrators for Decentralized Detection

Binary integrators are an important part of the receiver in many operating radar systems. The optimisation of a binary integrator is not a simple task, because it requires the solution of a (k x n)-dimensional nonlinear optimisation problem, where n is the number of integrated bits (or the number of sensors in a distributed radar or sensor network) and k is the number of the design parameters of the single-pulse detector. An algorithm that converts the multi-dimensional optimisation problem into a one-dimensional problem, so reducing considerably the computational complexity, is developed. This reduction in computational complexity makes the real-time optimisation possible and practical, so it is very helpful for mobile sites in which the optimisation should be performed continually. The proposed algorithm can be applied when either the 'AND' or the 'OR' integration rule is adopted. The results are illustrated by means of two study cases. In the first case, the binary integrator of a constant false alarm rate radar detector is optimised; in the second one a decentralised detection system composed by n similar sensors is considered and the decision rules are jointly optimised according to the Neyman-Pearson criterion.