GVGExp: Communication-Constrained Multi-Robot Exploration System based on Generalized Voronoi Graphs

Kizito Masaba, Alberto Quattrini Li

Abstract: This paper presents GVGExp, a recurrent-connectivity exploration strategy for multi-robot systems to discover unknown environments under communication-constrained conditions. A robust multi-robot exploration strategy with communication constraints is important to accomplish several applications, e.g., underwater or planetary exploration. Mainstream multi-robot exploration strategies have considered unlimited communication. In addition, these strategies delegate the resolution of potential path collisions to local planners. In this paper, we explicitly focus on minimizing the number of communication events between robots given the limited bandwidth that can be available in real missions, as well as explicitly minimizing potential path interference between robots. GVGExp incrementally builds a Generalized Voronoi Graph (GVG), which is used by the robots to determine the topology of the environment. We introduce a novel property of the GVG called gate to identify a region (subtree) that is uniquely assigned to each robot, with no overlap, thus minimizing potential path interference. Whenever a robot finishes exploring a region, in a depth-first search fashion, or finds a loop connecting to other parts of the environment, the robot shares information with other robots that are in range to enable coordination. We performed numerous simulations to evaluate our proposed strategy and compared it with other state-of-the-art methods. Experimental results show that GVGExp is able to explore the environments in a relatively short amount of time, significantly reducing communication events and path interference.