Controller design and real-time fault diagnosis for a humanoid robot

An effective controller is crucial for a humanoid robot since a humanoid robot generally has more than thirty DOFs to be controlled in real-time and needs to deal with information of multiple sensors. On the other hand, real-time fault diagnosis is increasingly important for the humanoid robot due to its mechanism and control complexity and inherent instability of risking tipping itself over. In this paper, we propose a distributed controller consisting of the online planning sub-system and the motion control sub-system based on CAN bus and Ethernet for humanoid robots. Moreover, a real-time fault diagnosis method is proposed to observe the most probable faults, such as joint over-limit, force/torque sensor failure, encoder failure, and inertial sensor failure. The effectiveness of our designed controller and fault diagnosis was confirmed by experiments on our newly-built humanoid robot.