Trajectory Prediction of UAV Swarm based on Neural Relational Inference Model without Physical Control Law

Abstract

Many Unmanned Aerial Vehicle(UAV) swarm can be treated as an interacting system because they have common behavior and dynamics. Extensive applications of UAV swarm in the local war recently have been successfully implemented into the air combat. Therefore, how to counter UAV swarm effectively has become an important approach to beat the enemy in the battles. The key to tackle the counter UAV swarm’ issue is to detect their precise position. And predicting the trajectory of UAV swarm effectively can help to improve the performance of position detection. In this paper, we apply the Neural Relational Inference(NRI) model in our framework. Furthermore, by the method of analyzing the dynamical equations that directly affect the trajectory of different interacting systems, we propose a Mapping Table to build the relationship between the spring / charged particles model and the UAV swarm. Based on the Mapping Table and NRI model, we propose a novel framework to predict the trajectory of UAV swarm. The experimental results show that our framework can effectively predict the trajectory of UAV swarm.