Spatial prediction of wireless links and its application to the path control of mobile robots

Abstract

We consider the path planning problem of a mobile robot that has to travel towards a given target location. The robot shares the environment with other mobile robots, altogether forming a wireless mobile ad hoc network relaying data in a multi-hop fashion. In this scenario, the robot's path planner has to optimally balance two potentially conflicting goals: keep the traveled distance within an assigned maximum value while letting the robot effectively communicate with the other robots in the network. We propose a solution method relying on the use of a link quality predictor built offline through a supervised learning approach. Together with the information gathered online from the other robots, the predictor allows to adaptively build a spatial map of expected communication quality, for both local and distant areas. In turn, the map is used by the path planner, based on a mixed integer linear formulation and an intelligent strategy for discretizing the environment, to iteratively find the best network-aware path to follow. The proposed approach is evaluated in various realistic simulation scenarios, showing the effectiveness of using the link quality map and the robustness to different restrictions regarding available information and computational resources.