Research On Legged Robot Motion Control Based On A Multilayer CPG

Legged robot is a mechanical system that has multiple degree of freedom, strongly nonlinear, and multi-redundant. Therefore, controlling the legged robot in unknown, complex environment is a difficult problem that is strongly emphasised on in the robot research field. In natural environment, central pattern generator(CPG) of animals can control animal motion in unknown and complex environment by utilizing multiple sensory feedbacks to control motion of animals in environment, and the coordinate motions between animal joints and legs, while the motion show strong stability, adaptability and coordination.Therefore, constructing motion control method of legged robot by simulating CPG control mechanism, is an effective way to solve legged robot control problem in unknown and complex environment. However, existing CPG bionic control methods lack the complete simulation of biological CPG control mechanisms, most of them only focus on interaction control between robot and environment, or on the coordination control between robot joints or legs, while in these researches, the speed of neuron response is slow, the interaction between robot and environment is hard to be regulated by sensory feedbacks, and it is hard for these CPG control mechanism to simultaneously coordinate multiple joints and legs, or adjust the coordination according to robot motion state. Therefore, it is hard for these control methods to have strong stability, adaptability and coordination in unknown and complex environments.We research on a multilayer CPG control method that can complete simulate the animal CPG motion control mechanism. In this article, the multilayer structure of animal neural network, interaction control mechanism between animal and environment, and the coordination control of multiple joints and legs are simultaneously simulated, a complete bionic control stucture which can control the motion of legged robot in complex environment independently is constructed, and has stability, adaptablity to environment changes. Specific studies include the following aspects:In response to the problems in existing CPG inter neuron such as the response speed is slow, and the output of motoneuron lacks online regulating, the inter neuron model is constructed ultilizing the Hopfield neuron model, and the quickly response of neuron is achieved; the model of motor neuron is also constructed simulating the working mechanism of biological neuron-muscle system, and regulation of robot joint in real time is achived. On this basis, against the problems in existing CPG control method, such as lacking of plentiful motion forms,and lacking of regulatory informations, the inter neuron model and the motor neuron model are combined, and a multilayer CPG neuron network model is proposed by simulating the CPG control mechanism of layered reception and execution of feedbacks and motion commands. Stable rhythmic motion control of robot joints are achieved, and by comparing experiments, the rapidity and stability is verified.A single-legged robot in the sagittal plane is chosen as research object, and the stable interaction control method for each degree of freedom is proposed by analysing the interaction between robot and environment. According to this control method, multiple feedbacks are introduced into the inter neuron layer and the motor neuron layer in the multilayer CPG neuron network, and the closed-loop and stable control between robot, CPG control mechanism and environment is realized by adjusting the motion rhythm, pattern and joint trajectory of robot with multiple feedbacks. We validated the stability and adaptability of the interaction control method on the single-legged robot platform, while comparing with biological experiments.Simulating biological coordination, A coordinated control method for multiple joints is proposed by adding interneuron delay sector and adjusting motor neurons of each joint with sensory feedbacks, by this method, the forward velocity and jumping height is increased. A coordinated control method that can simultaneously coordinate robot joints and legs, and adjusting the coordination according to the interaction state between robot and environment is also proposed on the basis of joints coordinate mechanism, and alternately motion of the two legs is realized by the coordination between robot legs. Combining the CPG interactive control mechanism and the coordination control mechanism, a multilayer CPG control method that can completely simulate biological is proposed, which can independently control the motion of biped robot in sagittal plane. The stability, adaptability and coordination of this control method is validitaed on simulation platform of multi-joint biped robot.Quadruped robot is selected as an application and experimental validation platform for multilayer CPG control method. A simplified control method for quadruped robot is proposed according to mechanical characteristics of quadruped robot. By simplifying the control mechanism of quadruped robot and adding the control mechanism of yaw and rotation, quadruped robot motion control is realized ultilizing the multilayer CPG control method. Stability, adaptability, coordination and precision of movement of this control method were verified and analyzed both on simulation platform and entity experiment platform. Experimental results show that this method effectively realized completely simulating biological CPG mechanism, and is able to be extended to the three-dimensional environment and controlling of multi-legged robot, can independently generate rhythmic movement of the robot in unknown, complex environment, while the movement show strong stability, coordination and adaptability.