Meta-heuristics for improved RF emitter localization

Abstract

Locating Radio Frequency (RF) emitters can be done with a number of methods, but cheap and widely available sensors make the Power Difference of Arrival (PDOA) technique a prominent choice. Predicting the location of an unknown RF emitter can be seen as a continuous optimization problem, minimizing the error w.r.t. the sensor measurements gathered. Most instances of this problem feature multi-modality, making these challenging to solve. This paper presents an analysis of the performance of evolutionary computation and other meta-heuristic methods on this real-world problem. We applied the Nelder-Mead method, Genetic Algorithm, Covariance Matrix Adaptation Evolutionary Strategies, Particle Swarm Optimization and Differential Evolution. The use of meta-heuristics solved the minimization problem more efficiently and precisely, compared to brute force search, potentially allowing for a more widespread use of the PDOA method. To compare algorithms two different metrics were proposed: average distance miss and median distance miss, giving insight into the algorithms’ performance. Finally, the use of an adaptive mutation step proved important.