A Distributed Spanning Tree Method for Extracting Systems and Environmental Information from a Network of Mobile Robots

A multi-robot system, like a robot formation, contains information that is distributed throughout the system. As the collective increases in numbers or explores distant or difficult areas, obtaining collective situational awareness becomes critical. We propose a method for extracting system and environmental information distributed over a collective of robots. Introduction and Background Endsley (1988) defines situational awareness (SA) as “the perception of the elements in the environment within a volume of space and time, comprehension of their meaning, and the projection of their status in the near future.” A robot in a collection of robots, which are coordinating as a swarm (c.f., McLurkin, 2004) or a formation (c.f., Fredslund & Mataric, 2002), can ascertain information about itself and its immediate surroundings. This individual robot SA is useful for tasks, such as maintaining relationships within the collective. However to accomplish group level tasks, such as the collective efficiently pathing around obstacles, the SA of the entire assemblage of robots is necessary; we term this collective situational awareness (CSA). CSA can be used to issue commands across the collective to react to the environment or change behaviors to carry out a task. It can also provide human operators with environmental information to create a comprehensive survey. CSA is attained through the accumulation of information embodied in individual SA, which is naturally distributed information across the collective. Previous work in CSA has focused on recognizing boundaries of the collective. Fekete et al. (2005) develops a heuristic to estimate the size of the boundary graphs based on the observation that interior boundaries are often smaller than outer boundary. McLurkin & Demaine (2009) implements A method for boundary detection where a robot determines a “local boundary classification” based on its sensor field-ofview—if no other units are located in this region, the robot is classified as a “boundary”; otherwise, the robot is classified as “non-boundary”. Our approach to CSA works with a formation of robots in which neighborhood relationships are established and maintained between two or more robots (Mead et al., 2009). Having this relationship allows the application of a spanning tree-based algorithm applied similarly to communication networks for optimizing message passing (Pendarkis, 2001).