Cooperative Path Planning for Multi-Robot Systems in Dynamic Domains

Abstract

A stationary robot in an assembly line expects to execute all its hard-coded movements without any collision. Its recognition is limited to its tasks and the corresponding area of operation. In case of a failure an alarm is given and it stops working until a human being corrects the fault. The assembly robot does not react independently, which is the exact opposite to autonomous mobile robots. An autonomous mobile robot has to solve many problems an assembly robot is never confronted with. While the latter moves according to its hard-coded patterns, a mobile robot on the contrary moves in a potentially unknown and dynamic environment. Operating in such environments requires the mobile robot to localise its position relative to the environment, sense its surroundings continuously, and move accordingly. In order to achieve its goals in a changing environment, a mobile robot also has to adapt its actions according to the changes. Hard-coded action sequences are too limited and error-prone to reliably achieve a specific goal. Finally, a secure execution demands the robot to plan its movements and their consequences. Collision-avoidance is an important topic in this context. The requirements for an autonomous mobile robot which is part of a team of robots are even more challenging. A robot has to recognise its team members separately, include their positions and abilities in its own planning and consider their paths for its own path planning. Nevertheless, many reasons for accepting this overhead exist. One robot handling complicated tasks is possibly more expensive and complex than several single robots. Furthermore, several robots are able to solve a given task faster than a single robot, by dividing the task into subtasks and assigning them to different robots. Moreover, a team of robots exhibits reliability, as malfunctioning robots can be replaced by team members at runtime. Finally, a large spectrum of tasks can be covered by a group of robots at once, while a single robot has to be adapted for each specific task. A central requirement to benefit from these advantages is the ability to coordinate multiple autonomous mobile robots. For this chapter we will confine ourselves to the problem of path planning in a team of autonomous mobile robots. Refining the requirements, we assume the following conditions: