Study Of Real-time Humanoid Robot Motion Planning Method

Recently, the research on humanoid robot has developed rapidly, and a large number ofachievements have been made in this field. As the basic character of human beings, stabledynamically walking has become a focus in robot research field. This paper mainly focuses onthe researchinto motionplanningandcontrollingon-lineofhumanoidrobot.Firstly, the characterand mainframe parameters of the humanoid robot areintroduced, andwith the consideration of real-time feature, reliabilityand expansibilityand so on, the feasibilityofapplicationofWindowsCE.NEToperationsystemonhumanoidrobotplatformis analyzed.Secondly, the humanoid robot platform is simplified, and on the basis of ignoring thecoupling effect between the frontal motion and the lateral motion, the frontal and the lateraldynamics models of humanoid robot are built up using Lagrange method. Subsequent work ismeasuringforceandtorqueinformationbetween the feetofhumanoidrobotandgroundwiththesix-dimension force/torque sensors, and formularizing the calculation expression of realZMP(Zero Moment Point) in single supporting phase and double supporting phase, which isusedascriteriaofstabilitythroughtheactualmotionofhumanoidrobot.Thirdly, a feasible method to solve the inverse-kinemics problem based on CMAC isrealized and validated by simulation results under Matlab. This method use CMAC NeuralNetwork to approximate to the inverse-kinematics model of humanoid robot, and gets themapping relationship between stance warp and angle adjustment value of every joint. So, theproblems ofrequiringa mass ofcomputationand thesolutionmaybenot unique,whichface thetraditionalanalyticalway,areavoidedwiththismethod.In the end, the gait is planned on the basis of the ZMP stability criteria and the CMACbased inverse-kinematics solution method, and with the information detected by force/torquesensors, the algorithm for gait control of humanoid robot in real time is discussed andresearched.