Abstract: The design and control of elbow joint are the research priorities of humanoid arm. To obtain the elbow joint model, a test platform is constructed for static modeling of a single pneumatic muscle,and then a nominal model of cascade elbow joint is established. The least square parameter identification is used to obtain the parameters of model. A sliding mode control law is designed based on Luenberger disturbance observer. The elbow joint location tracking with proportion integration differentiation (PID) control, sliding mode control (SMC) and sliding mode control based on disturbance observer (SMCDO) are simulated. Three control algorithms are tested by taking half a bottle of mineral water held at the end of humanoid arm as a load and uncertain outside disturbance. The simulated and experimental results show that the location tracking accuracy and robustness of sliding mode control based on disturbance observer are better than those of sliding mode control and PID control. Key

Key words: controlscienceandtechnology, pneumaticmuscle, elbowjoint, slidingmodecontrol, disturbanceobserver