Research On Intelligent Learning Motion Control Of Biped Omni-directional Walking For Gorilla Robot

In recent years, scholars have done a lot of research on biped walking robot inthe aspect of strengthening environmental adaptation, and the environmentaladaptability of the walking robot has been improved. However, the research onbiped walking robot is still in its infancy. The research on adaptive intelligentcontrol is in the theory stage, which is used in the actual intelligent control isnumbered. In the future, it is extremely important for intelligent control toresearch intelligent learning. Intelligent robot is the best way to improve the abilityof the biped walking robot to adapt to the environment.This paper focuses on the omni-directional walking planning and walkinglearning research to provide a method to improve the flexibility and adaptability ofthe bipedal walking robot. Concretely speaking, firstly,Fourier series approximationplanning method for the omni-directional walking has been introduced. The methodwhich makes full use of characteristics of the walking cycle, and combines with thestability properties of3-D inverted pendulum based on ZMP is uesd to plan thecomplex motion of mass center, and provides a theoretical basis for the stability ofwalking robot. In terms of.programming and implementation, the method is simpleand easy to complete by useing the series approximation method, thus, it is an idealwalking planning method. The trajectorie of COG for turining is planned bybuilding the relation between forward walking and turning walking, and the planingof omni-directional walking is completed for the bipedal walking robot.The research on walk learning is carried out by learning success walkingsamples obtained by using the above method. The main information which affectsthe walking stability is learned by using walk learning. There are three differentsituations for walk learning including different step length, different step velocity,different step length and velocity and different turning radius, and they are analyzedby applying on a bipedal robot. The unexperienced walking samples are obtainedby learning of CMAC, and are valued by analysing relation between the joint anglesand simulation. Finally, a method abouting adjust the weights of CMAC for theCMAC-PD control system is given. It can realize the online learning and real-timecontrol by using CMAC walking learning and control, thus, it is a better mean toachieve adaptive intelligent robot control.