Research On The Control Algorithm Of A Cheetah Robot's Galloping Based On A Dynamically Equivalent Model

Analyzing and controlling the running of a quadruped is difficult,especially for a cheetah's galloping,since it is complex both temporally and spatially because of the multiple degrees of freedom,the effects of the rhythms and the dynamics and the interactions between them.The research is intended to establish a dynamically equivalent model for the controlled object and then establish a control method for application.The research consists of four parts:Firstly,a rhythm generator is established based on finite state machines and integral rules to describe the running sequence of a cheetah.The rhythm generator is based on the characteristics of the cheetah's galloping rhythm.Finite state machines regulate the transitions between the motion phases and the integral rules regulate the variation of the phase positions.Phases and phase positions are used to indicate motion stages and moments.All the rhythm parameters vary automatically,and the variations of the parameters are adjusted by the kinetic perception so that the rhythm signals synchronize with the motion states.Secondly,a dynamically equivalent model of a cheetah's galloping is established to analyze the factors which influence the kinematics characteristics of the torso.In the dynamically equivalent model,the inertia characters are defined by a two-mass-point system,and the displacements of the feet relative to the center of the mass are dominated by the action of the linkages.The influences of the action of the linkages on the kinematics characteristics of the center of mass are analyzed and they regulate the planning of the foot trajectories.To make the galloping sequential,the control during the supporting phases should be emphasized to make an expected ballistic flight when the feet are all off the ground.Then an algorithm which uses a feedforward control as a main way and a backward control as an auxiliary is established for the robot's real-time control.During the leg's swing phase,a PID method is used to track the basic foot trajectories;during the leg's stance phase,the feedings of the foot location are generated by the feedforward and feedback control method.The foot feedings during the stance phase are the emphasis of planning: the expected kinematics characteristics of the center of mass together with the foot locations generate the expected kinematics characteristics of the feet which are used as input to the feedforward control;the compensations are acquired from the error between the expected and actual kinematics characteristics and they are used as input to the feedback control.Finally,the galloping in the simulation and the physical prototype experiment are tested to verify the algorithm.The torso structure is designed to increase the stride distance.The control algorithm and the virtual prototype are established in the Simulink-RecurDyn joint simulation and the virtual prototype forms a stable periodic galloping.SQBot II is an improved version of SQBot I and is designed for a higher running speed,while the former running platform is improved.SQBot II gallops steadily on the platform.The result shows that the algorithm can make a stable periodic galloping in both the simulation and the experiment,and it can tolerate the initial altitude error so it is robust in some way.The rhythm generator and the dynamically equivalent model in the algorithm have wide applicability for a quadruped's running.