SResearch On Humanoid Robot Head And Its Dynamic Control

Humanoid robot is one of the most precessional subjects in the robot research field, which integrates mechanism, electronics, computer, material, sensors, control technology, artificial intelligence. It reflects the develop level of science and technology of a country. As the self-subject"Robonaut"of the State Key Laboratory of Robotics and System, the humanoid robot head is developed in this paper, and the dynamic control is studied. This system will support the Active Vision for the"Robonaut".First, according to the requirement of visaul measurement and control, the design principle and performance parameter is proposed. On this foundation, the mechanical system is designed. The neck base joint is designed to achieve pitching and swing movement, using bevel gear differential mechanism. To achieve mechatronics, the separated sensor board, driver board and control board based on FPGA are designed. A distributed control architecture based on vision processing-central control-joint control is adopted. The whole head system is 1.3 kilograms, with 3 Degree Of Freedom (DOF) (pitching, swing and rolling), compact structure and humanoid appearance.Accomplished the design of the system, the models of stereo vision, kinematics and dynamics is studied. The stereo vision model is constructed by vision disparity elements. Combined with direct kinematics, the conversion from image-coordinate to world-coordinate is achieved. In inverse kinematics, a novel D-H method based on virtual arm is proposed, bring the speacial head inverse kinematics to the general D-H method. The robot link dynamic model and torque conversion model are constructed.Contrapose to the error and uncertainty of the model, Sliding Mode Control (SMC) strategy is used for its robust by the system uncertainties and outside disturbance. Contrapose to the chattering of SMC, the bating arithmetic is the key study in the paper. An improved exponent reaching law solution and two composite control based on SMC are proposed, and simulated through Matlab/Simulink. The simulation indicates that the improved exponent reaching law solution improves the reaching quality, minishes the error and the two composite control can bate the chattering effectivelyIn the end, the joint position control experiment and active vision tracking experiment are accomplished on the real humanoid head system. The experiments indicate: The improved exponent reaching law solution and two composite control based on SMC can bate the chattering effectively. And the system can track the pingpang ball in a velocity range using Position-based vison servo strategy.