Decomposition Optimization Algorithms for Distributed Radar Systems

Distributed radar systems are capable of enhancing the detection performance by using multiple widely spaced distributed antennas. With prior statistic information of targets, resource allocation is of critical importance for further improving the system's achievable performance. In this paper, the total transmitted power is minimized at a given mean-square target-estimation error. We derive two iterative decomposition algorithms for solving this nonconvex constrained optimization problem, namely, the optimality condition decomposition (OCD)-based and the alternating direction method of multipliers (ADMM)-based algorithms. Both the convergence performance and the computational complexity of our algorithms are analyzed theoretically, which are then confirmed by our simulation results. The OCD method imposes a much lower computational burden per iteration, while the ADMM method exhibits a higher per-iteration complexity, but as a benefit of its significantly faster convergence speed, it requires less iterations. Therefore, the ADMM imposes a lower total complexity than the OCD. The results also show that both of our schemes outperform the state-of-the-art benchmark scheme for the multiple-target case, in terms of the total power allocated, at the cost of some degradation in localization accuracy. For the single-target case, all the three algorithms achieve similar performance. Our ADMM algorithm has similar total computational complexity per iteration and convergence speed to those of the benchmark.