Multisensor Management Algorithm for Airborne Sensors Using Frank-Wolfe Method

TL;DR

This paper introduces a multisensor management algorithm for airborne target tracking using the Frank-Wolfe optimization method, which improves sensor deployment and guidance for better tracking performance.

Contribution

It develops a novel Frank-Wolfe based algorithm for optimal sensor configuration and guidance in airborne multisensor target tracking, ensuring feasible solutions and enhanced performance.

Findings

The algorithm outperforms traditional methods in simulations.

It guarantees sensors remain within feasible deployment points.

The method effectively optimizes sensor guidance for dynamic targets.

Abstract

This study proposes an airborne multisensor management algorithm for target tracking, taking each of multiple unmanned aircraft as a sensor. The purpose of the algorithm is to determine the configuration of the sensor deployment and to guide the mobile sensors to track moving targets in an optimal way. The cost function as a performance metric is defined as a combination of the D-optimality criterion of the Fisher information matrix. The convexity of the cost function is proved and the optimal solution for deployment and guidance problem is derived by the Frank-Wolfe method, also known as the conditional gradient descent method. An intuitive optimal approach to deal with the problem is to direct the sensor to the optimal position obtained by solving a nonlinear optimization problem. On the other hand, the proposed method takes the conditional gradient of the cost function as the command to the deployed sensors, so that the sensors are guaranteed to be in the feasible points and they achieve the current best performance. Simulation results demonstrate that the proposed algorithm provides better performance than directing each sensor to its optimal position.