Research On Gait And Motion Control Of Quadruped Robot

After hundreds of millions of years of natural selection,mammals have evolved the best bone parameters and the best exercise performance.For thousands of years,imitating nature has become the research direction of human beings.From the early“wooden ox flowing horse” to the present “walking car”,all embody the crystallization of human wisdom.At present,extracting biological structural parameters,imitating biological motion patterns and mining biological neural control mode are important parts of the combination of bionics and mechanism.As a nonlinear and complex multibody dynamic system,quadruped robot is not only involved in bionics,mechanism,control theory and other interdisciplinary subjects,but also has strong correlation and coupling between the bodies,which is a great challenge to its research.Therefore,many basic theories and key technologies need to be further studied.This paper takes the stability and environmental adaptability of quadruped robot as the research goal,and studies the key technologies such as structure design,trajectory planning,gait control mode,etc.,which provides a theoretical basis for the gait and motion control of quadruped robot.The specific work is as follows:(1)According to the theory of biological anatomy,it is found that dogs walk in the way of plantar movement,which has high-speed movement performance.This thesis takes dogs as bionic objects,obtains the simplest form of animal movement,determines the design goal and completes the design and processing of the whole quadruped robot.According to the D-H method,the forward and inverse kinematics equations of the simplified single leg are obtained,and the inverse kinematics equations of the real quadruped robot are obtained by kinematics equivalence.The workspace of the robot's foot end is analyzed by Monte Carlo method within the range of joint angle.(2)The existing trajectory planning methods lack of bionic characteristics and do not conform to the laws of biological motion.In view of the above problems,this thesis proposes a quintic polynomial foot trajectory method based on zero impact at the foot end.The backward swing section and retraction section are added to the horizontal trajectory segment to make it more bionic.The second derivative of quintic polynomial is continuous,and its acceleration curve is no longer a straight line,which is an ideal foot trajectory.The single leg control system is built by Solid Works-Sim Mechanics cosimulation platform,and the motion parameters of single leg are obtained.(3)In order to reduce energy consumption and improve walking efficiency,robots should choose the appropriate speed according to terrain conditions like mammals,which involves gait switching problem.At present,few people explore the gait switching between trajectory and trajectory.In this paper,the zero impact quintic polynomial of foot end is selected as the foot trajectory,and the large period joint driving angle and delay time are introduced into the transition section.The variable period control of the transition section is proposed,which realizes the smooth transition of the driving joint angle,and solves the stability problem of the robot through the trajectory trajectory step state switching.Under the idea of variable period control in the transition period,the constant speed control and variable step control are used to complete the transition from walk gait to trot gait.(4)The CPG network topology is built by using the mathematical model of Hopf oscillator.The influence of parameters on the output curve of Hopf oscillator is obtained,and the independent adjustment performance of Hopf oscillator parameters is determined,which lays the foundation for the independent adjustment of hip and knee joint drive curve.The output harmonic signal of Hopf oscillator can achieve good control effect on double beat gait,but the output curve of oscillator can not be used for four beat gait because of the time inconsistency between the support phase and the swing phase due to the change of duty cycle.In this paper,the swing phase frequency is used to change the support phase and swing phase time,and the duty cycle is adjusted to achieve the desired trajectory output.Through ADAMS-Simulink co-simulation platform,walk gait and trot gait are realized.In order to realize the transition from one gait to another,the matrix of the gait is often replaced directly.This method will lead to the problems of vibration stop,phase lock and sharp angle in the angular drive curve of robot joints,which will lead to the sudden change of velocity and acceleration.In this paper,the segmentation function is introduced into CPG model,and a joint algorithm of Hopf oscillator and segment function is constructed.The duty cycle of the robot is changed by changing the phase difference of the right rear leg in real time,so as to realize the smooth transition of joint driving curve and ensure the stable walking of the robot.The comparison between the CPG based gait generation method and the modelbased gait generation method proves that the CPG control method has higher stability.In order to combine the advantages of CPG neural control mode and foot trajectory planning,neural network fitting function is used to realize the nonlinear mapping between CPG model and foot trajectory model.(5)In order to verify the effectiveness of the algorithm,a quadruped robot named KUST-canine is designed on the basis of kinematics and bionics research.Based on the existing conditions in the laboratory,the single leg motion experiment of the quadruped robot is carried out,in order to obtain the motion trajectory of the robot leg,the laser tracker is used to collect the foot endpoint set in real time,and the fitting function of SA software is used to obtain the foot trajectory.The experimental results show that the robot can move according to the expected trajectory,and achieve zero impact at the foot end.On the basis of single leg motion control,the walk gait experiment and trot gait experiment are carried out to prove the effectiveness of the proposed algorithm.