Humanoid Robot Control System Design And Attitude Control

Humanoid robot has been well recevied as one of the most active research topics in the robotics community as well as the long term ambition of technology development. The advancement on humanoid robots represents the state-of-the-art of a nation’s technological and scientific development. Featured with its human-like outlook, humanoid robots have a promising future in replacing human power from dangerous and heavy labor work, thus improving the productivity and effeciency of human life. With functional humanoid robots serving around, human can be more focused on high-level intelligent and creational activities. It is forseeable that robotics technology, especially humanoid robots, will play a key role in shaping the future society and the world in the years to come.The main contents of this thesis includes:1. Targeting at the capability of fast locomotion and manipulation, a novel network structure for multi-axis motion control based on industrial real-time ethernet standard EPA is proposed. The proposed design improves the bandwith and real-time performance of the distributed motion control system used on humanoid robot Kong-II;2. A software framework that supports humanoid robots executing fast manipulation and biped walking tasks is proposed. The framework, named as EvolvingMind, adopts multi-layer design and is featured with a modular structure which effectively reduces coupling between software and hardware and seperates implementation and algorithm. The adoptation of the framework reduces the cost of function upgrading and maintenance of the software system;3. Basing on experimental observations, main disturbances and failing factors of biped walking is discussed. We vision the landing force impact endangers the stability of the humanoid robot, and proposed an impedance control based approach to reduce the stiffness of the swing leg so that soft and firm landing can be achieved. We also attacked the problem from the energy point of view, and proposed a passivity-based approach to improve the overall stability of the robot by regulating the input energy from landing impact. A least square based angular momentum control method is also addressed to compensate the momentum generated during biped walking, which is generally ignored in most motion planning schemes;4. Challenged by the large reaction force caused from acceleration in arms during fast manipulation tasks, an impedance based standing pose compenstaion method is proposed, implemented and tested. Using different spirng-damper model for different leg, the robot’s dynamic performance against disturbances is improved as well as the pose stability strenghthened. A momentum optimization method is also studied to reduce the angular momentum cased by fast arm motions.Based on the system, software and algorithms introduced in this thesis, the humanoid robot Kong-Ⅰ and Kong-Ⅱ developed by the author’s research group is now capable to play table tennis against a human player of another humanoid robot. The best record of human-robot table tennis rally reaches 144 turns, and that of robot-robot rally reaches 114, which verifies the effectiveness and reliability of this research. The humanoid robot can also achieve biped walking with a top speed of 1.07km/h. These experiments prove the effectiveness and reliability of the proposed approach.At last of this thesis, we conclude the main content and address the future plan of our research.