Dynamic skill-based software architecture for locomotion control in shape reconfigurable mobile robots

This paper presents dynamic skill-based software architecture for locomotion control in shape reconfigurable mobile robots. The paper first describes multi-thread based synchronization locomotion control model, which is applied in a mobile snake robot and physically reconfigured four-legged robot. The robots share common multithreaded software architecture and hardware modules, which use the concepts of sections and units. The various "skill sets" are mapped to various threads and these threads again mapped to section modules of the robots. Each robot section consists of smallest movable robot modules termed as units. The units comprise of sensors and actuator blocks. We map these units to "skills" in our architecture. The synchronized sets of skills are responsible for robot locomotion. The dynamic selection and synchronization of "skill-sets" enable the mobile robot to perform many tasks in complex situations. The architecture is a layered architecture which includes reasoning layer, skill-set layer, skill layer and sensor-actuator layer. The architecture can be dynamically reconfigured for various movements like backward, forward, side winding etc using proper selection of skill-sets