Research On Self-tuning Fuzzy Control Of The Humanoid Robot Based On Walking Error

Humanoid robot is one of the most representative intelligent robot types. Research in this field has become the hotspot of robotics technology.The dissertation summarizes the research development of the technologies and theories of the humanoid robot stable walking. Aimed at the nonlinear servo control of robot joint, stability of robot walking and swinging foot's landing impact force, develop the research on self-tuning fuzzy control of the humanoid robot based on walking error, and have completed technology research and validation on the THBIP-I prototype platform of Tsinghua University.Major achievements of this dissertation are as following:Analyzes the servo control performance of DOFs (degree of freedom) of ankle joint, which is affected by the nonlinear transmission ratios of the bar linkages. Therefore based on the joint position error and the change of the ratio, designs the humanoid robot's fuzzy joint controller with self-tuning factor. The experiments of ankle joint prove that the controller possesses well servo dynamic performance of humanoid robot joint, and satisfies the variational stiffness performance of ankle joint. Hence, the joint servo problem in humanoid stable walk has been solved effectively. Deduces equations of desired ZMP (Zero Moment Point) with joint generalized coordinate, and establishes ground reaction force measure system with Universal Force/Moment sensor, then simplifies calculating model of actual ZMP. Analyzes the joint movement contributions to ZMP and the reason for ZMP error and the tipping moment acted on the robot, then based on ZMP error, presents the compensate control strategy for ankle joint. The strategy utilizes with ZMP error and its change, compensates the motion control commands of pitch and roll DOFs of support ankle joint online. Experiments of support ankle joints indicate that the compensate controller effectively reduces ZMP error and the swing of actual ZMP, and the walking stability of prototype is enhanced.Deduces the impact dynamic equations at the landing time of humanoid robot's swing foot, and analyzes the reason for impact caused by the swing foot motion error. Then, based on the consideration of decreasing impact force by correcting the robot swing foot motion error, develop a fuzzy compensate control strategy for swing foot landing. The motion control commands of pitch DOFs of hip joint, knee joint, and ankle joint of swing leg are gradually compensated online. Experiments of swing leg prove that the controller effectively improves pace and swing foot height before landing, the actual landing time tends to the desired landing time, and the impact force between swing foot and ground is decreased, the change of which becomes smooth. As a result, the vibration of prototype walking is decreased.Following the principle of "precision increased by degrees goes with intelligent descending" in intelligent control and based on CAN bus, designs a hierarchical distributed control system of the humanoid robot. The whole control system consists of an organization layer, which takes charge of task planning, system supervising; a coordination layer, which coordinates and distributes tasks; and an execution layer, which includes sensor system and joint servo control system. The functionality of coordination layer and execution layer has been verified in humanoid robot prototype's static walking experiment.