A New DNN Technique for Improving Relative Localization Accuracy Based on Distance Between Unmanned Swarm Nodes

Abstract

In an environment where multiple nodes move without fixed points, such as in swarms of unmanned robots, understanding the relative positions of the robots is crucial. This paper proposes a new relative positioning technology for understanding the relative positions of nodes in environments where GPS usage is difficult, such as indoors. Specifically, the proposed technology is a new deep neural network (DNN) technique that performs relative positioning using distance information between nodes. This paper proposes two new methods to enhance the performance of relative positioning based on existing DNN techniques. The first method ensures a minimum distance between reference nodes, and the second method involves selecting the optimal reference node. Through computer simulations, it was confirmed that coordinate estimation performance improves when a minimum distance between reference nodes is maintained. Based on these results, the method for selecting the optimal reference node was developed to choose nodes with greater distances between them. Using this method increases the accuracy of coordinate estimation compared to existing methods.