Dynamical Stability And Control Of Running Gaits Of Quadruped Robots

The studies of quadruped robot are inspired by the quadruped mammals.Quadruped robot is an amazing product of the combination of bionics and robotics.Throughout the history of the quadruped robot development,it is always the goal for the researchers to achieve high-speed,stable and flexible moving ability like the quadruped mammals in artificial counterparts.However,dynamics and motion control of the quadruped robot is very difficult because of its characteristics of high dimension,under actuated,variable topology,impact and model uncertainty.Dynamics and control of quadruped robot is one of the difficult and hot spots in quadruped robot research.In this dissertation,dynamics of the quadruped running gait in sagittal plane is discussed based on the simplified quadruped running model,the intrinsic dynamics and stability of the running gait are revealed using the method of Poincare mapping and transverse dynamics.Then,control strategies based on virtual constraints and learning algorithm are proposed for quadruped running.Research achievements in this dissertation provide a solid theoretical basis and feasible technical approach to practical applications of the quadruped robot.Considering the state switching and topology changes in the advancing process of the gait,the whole model is built using the method of hybrid dynamics.The continuity of this class of multiple subsystems running gait on the initial value are proved,which is the prerequisite and basis for subsequent dynamic analysis.Through the establishment of gait Poincare mapping,fixed points for gait cycle movement are calculated.We analyze the gait parameters interdependence,such as stride length,stride frequency and velocity,and the coupling relationship between legs stiffness,foot force and kinetic parameters of the quadruped.Thus,behavior mechanisms of running gait are revealed,which provides valuable reference for the design and control of quadruped robots.Furthermore,the robot running gait is linearized along the periodic orbit using the method of transverse dynamics.The quantitative analysis of stability of motion region around the periodic orbit is conducted by using the Lyapunov stability theory.On the basis of the above dynamics analysis,this dissertation presents control methods for two typical quadruped robot running gaits,i.e.,the bounding gait and galloping gait.Control strategy based on virtual constraint is put forward for bounding gait,the periodic trajectory are parameterized and converted to motion constraints to the leg joints,the constraints are applied to the robot motion and drive the robot convergent to the target gait.The control strategy realized the dimensionless Froude running speed of 1.01.Aiming at the complexity of galloping gait sequence,gallop motion control is transferred into the determination of 3 gait parameters using the motion characteristics of the gait.Then the locally weighted projection regression(LWPR)learning algorithm is adopted for learning different parameters under different speeds,realizes the Froude running speed of 0.84.Finally,this paper uses the principle of bionics to design a quadruped robot experimental platform Agi Dog to validate the proposed control algorithms.The experimental results show the effectiveness of the proposed method.The results of this study on robot dynamic characteristics and gait control provide not only the theoretical support,but also the feasible control strategy for the quadruped robot coming to utility,which shows the theoretical significance and practical value of this study.