Relative Localisation Algorithm for Neighbour Classification in Ad Hoc Networks of Moving Robots

This paper proposes a solution for mitigating mobility induced challenges in wireless ad hoc networks of moving robots operating in challenging environments. We consider a case of a network of moving robots instructed to share large data volumes in an environment characterized with lack of network infrastructure and uncertainty of signal measurements. The solution presented in this paper is based on estimating nodes' location and their direction of movement, using only occasional exchange of short messages. The first part of the solution presented in this paper is a new Relative Neighbour Localisation (RNL) algorithm. This algorithm uses random forest regression to estimate location and angle of movement based on signal strength measurements and exchange of angle and velocity information between nodes. The location information is then used in the new Neighbouring Source Selection (NSS) algorithm, which identifies the optimal neighbour data source using node's potentials as data sources, based on the nodes' locations and directions of movement. This enables the nodes to determine optimal data sources among the neighbouring nodes, thus helping to address frequent topology changes and uncertainty of communication channels, typical for network communication in extreme and challenging environments. The performance of presented solutions is evaluated using simulation and experimental testbed using six moving robot nodes equipped with WiFi-based networking interfaces. The results presented in this paper show the benefit of location estimation in improving data transfer efficiency, compared to the traditional use of only the strength of the received signals.