A new task-based control architecture for personal robots

Abstract

As the possible applications for personal robots increase and involve more complicated environments and tasks, they need more appropriate control architectures. Previous control architectures have fixed configurations that are optimized for specific applications or with limited flexibility. Consequently, they fail to provide the flexibility needed for various robot kinematical configurations as well as various tasks. To this end, we propose a new task-based control architecture called "supervised hybrid architecture (SHA)". SHA is based on supervised organization and distributed arbitration of hybrid controls of reconfigurable deliberative and reactive modules. It is composed of upper level hybrid control for high-level intelligence to interact with human and to plan tasks, as well as lower level hybrid control to allow low-level intelligence for prompt reaction in each robot configuration module. Through these double layers of the hybrid controller, we could provide the flexibility needed for so many different kinematical configurations and tasks. In addition, it is very easy to add or remove robot configuration modules. The proposed architecture are implemented and tested to show how it works successfully.