Design And Research Of Control Algorithm For Passive Biped Robot In Multi-walking Environment

The biped robot has a similar physical structure to human beings and can replace human beings to work in some repetitive and dangerous environments.However,the robot is less robust in the real walking environment,so achieving stable and efficient humanoid natural gait is an important research goal in the field of biped robots.Compared with the active control robot,the biped passive robot can make full use of its own intrinsic dynamic characteristics to improve control efficiency and reduce energy consumption without adding drivers to each joint.However,since the biped passive robot is very sensitive to external disturbances,some parameters change will cause non-periodic unstable gaits such as chaos.Therefore,in order to realize the stable walking of the biped passive robot under the realistic uneven road surface,this paper analyzes the dynamic characteristics of the robot under different slopes,and proposes effective control strategies for different walking environments,so as to improve the the stability and robustness of the passive robot walking on multi-walking road surface.The main research work of this paper is as follows:Firstly,based on the walking mechanism of humanoid walking robot,the hybrid dynamic equations of biped passive robots are established by using the Lagrange method and the conservation of angular momentum.Converting the nonlinear continuous dynamic system of robot walking into a discrete system through Poincaré mapping,and the asymptotic stability of the gait limit cycle of biped passive robots is analyzed.On this basis,aiming at the unstable gait of the robot caused by the change of the slope angle,analyze the energy conversion relationship of the biped passive robot during the swing and collision phase in a gait cycle,and the gait characteristics of the passive robot are obtained when the slope changes.Secondly,the existing control methods for biped robot under random roads mostly compensate the robots after interference,they can not adjust the walking gait of the robot timely and effectively before and after the slope change.If the road gradient can beknown in advance,more efficient control schemes can be designed for different pavement environments.Therefore,this paper proposes a random slope angle identification method based on GDFNN,which can realize the accurate identification of the slope angle within a certain range by training and learning of a large number of sample data.Providing support for the following stability control of the biped passive robot in the multi-walking environment.Thirdly,aiming at the problem of poor robustness of the simple passive robot in multi-walking environment,an improved spring-damped passive robot model are proposed.Analyze the energy variation of the biped robot in multiple walking environments such as uphill and downhill and design corresponding control methods for improved models walking under different slopes,so that the biped robot can make full use of the dynamic characteristics of the passive walking robot and maintain stable and efficient walking in multi-gait environment.Finally,the actual walking pavement environment includes different gradients of uphill and downhill and flat land.Aiming at the problem that the robot can not achieve timely and effective natural gait transformation when the slope changes,a deep enhancement learning method(DQN)based on RBF neural network and Q learning is proposed.With the change of the current state of the robot,the network weight and gait parameters update in real time through continuous training and learning,and choose the optimal gait through the optimal strategy,so that the robot can learn to adjust the real-time attitude on the uneven road surface.The effectiveness of the proposed control method is verified by system simulation.