Artificial immunity inspired cooperative failure recovery framework for mobile multi-robot system

Abstract

Robots are sophisticated machines which are specially designed to have the capabilities to handle operations, on behalf of human, in many different scenarios. In the past decades, the design of robot systems has been evolving and there are increasing numbers of possible applications of robot. Some systems can even be able to overcome the individual limitations and handle complex problems by combining the strengths of multiple robots.

To reduce the risk of human life, robots are now being put into missions under extremely dangerous or hazardous environment where human intervention is not tolerable , such as search-and-rescue missions inside damaged buildings after natural disasters and cleaning up of radioactive materials in nuclear accidence. Even though robots are dispensable, if they are damaged, disabled or trapped, the mission would not be accomplished. Therefore, the longevity of a robot system is always a challenge for robotic operations in such difficult environments.

To tackle this challenge, many studies focus on improving the design of individual robot, minimizing the chance of robot failure; or the way that how functioning robots may share the job of the failed robots. The way that how other robots can help failed robots recover, however, has yet to be widely discussed.

This thesis studies the feasibility of having multi-robot system with different automatic cooperative recovery abilities on top of its primary functions. A novel cooperative recovery framework is proposed for generic control among system primary functions and recovery behaviours. A number of experiments have been done to study the influence of cooperative recovery on a multi-robot system and how it can affect the system in terms of system performance, sustainability and overhead. An Immunity-based cooperative recovery model has also been created to overcome the drawback introduced by cooperative recovery, finding a balance between the two system objective among system productivity and longevity.

Two modified versions of cooperative recovery model are also included in this study to further maximize the system potential.