| Bionic quadruped robot has become a special robot with broad application prospects because of its excellent motion ability and strong terrain adaptability,and has been used in exploration,search and rescue and military fields.How to achieve energy-efficient and stable motion of quadruped robots has always been a research hotspot for scholars,which requires the robot to have a mechanical structure that stores energy and a dynamic control strategy with a feedback mechanism to maintain stability in various complex terrains.According to this research direction,this paper studies the energy-saving structure of leg,motion control and planning of quadruped robots,including structural design and system modeling,motion control,planning and state estimation,as well as simulation and physical testing.Firstly,the ontology structure of the robot was designed and the system mode was established.The legs of the robot were adopted a closed-loop four-bar linkage mechanism,and the knee joint is equipped with a controllable torsion spring based on the design principle of low energy consumption.The joint torque be provided by the spring during the support phase according to the energy consumption model,which can save the output energy of the knee joint motor.In order to describe the mathematical relationship between the joint angle and the foottip,the geometric method is used to deduce the forward and inverse kinematics,which lays a theoretical foundation for the subsequent foot-tip force optimization and energy consumption experiments.Secondly,the motion control algorithm of the robot is established.As a tool for human interacts with the environment,the controller of the robot should have high stability and be able to adapt to various complex environments.In the support phase,the body of the quadruped robot is assumed to be a single rigid body model(SRBD),and 12 state variables including the position,velocity,attitude angle and angular velocity are defined to fully describe the motion state of the quadruped robot.The central pattern generator(CPG)is used as the generation and conversion signal of the motion.The optimized foot force was solved through the model prediction principle,which is directly applied to the robot as an output to control the position and attitude of the robot.Thirdly the motion planning algorithm of the robot was established.In order to describe the state parameters such as the position of the robot,a single rigid body multi-point support model is constructed,and the precise estimation of the position and velocity of the center of the robot is realized by fusing the information of the foot velocity,position and body acceleration.The trajectory of the foot-tips was planning with signal coupled by the CPG and the desired motion parameters to reduce the excessive impact and avoid unnecessary energy loss.The trajectory is driven by the rhythm signal CPG,and various gaits can be realized by changing the parameters such as oscillation frequency.Finally,the simulation and physical testing of the robot were completed.A neural controller that combines MPC and CPG was proposed,a series of experiments including the trotting gait compound motion and walking gait motion were completed on the Adams platform,and the attitude adjustment,height adjustment experiments,forward experiments of various gaits including trotting,the static and dynamic energy consumption experiments were completed on the physical prototype.It can be seen that the energy saving of the leg structure and the responsiveness of the controller meet the design requirements by analyzing the simulation results and platform experimental results. |
