Corridors for robot team navigation

Abstract

In previous work, we proposed the use of virtual corridors to move a group of agents (animals) from one location to another. Using virtual fence technology, we can create a corridor from a set of fences to generate the desired gross motion of the herd without overspecifying paths for the animals. This allows us to reduce the amount of stimulus given to the animals while allowing them to naturally avoid obstacles (or other agents) without having to consider them during planning. In addition, a single corridor can be used by the entire group, simplifying planning and requiring only limited communication. In this work, we have implemented virtual corridors for robot team navigation, both in simulation and on two Pioneer mobile robots, and performed a number of experiments with two aims in mind. First, we wish to determine if, and under what circumstances, corridors are a practical solution to navigation by a team of robots. Secondly, by investigating different underlying motion algorithms for the robots, we may be able to suggest how successful corridors can be for different systems (including animals). We show success in open environments over a variety of experiments, with varying efficiency from the different motion models, and partial success among complex obstacles and with teams of up to five robots using a single corridor simultaneously.