Motion Plan And Coordinated Control Of A Quadruped Robot

Compared with the biped, the quadruped has better stability and larger capacity of load, also the quadruped has simpler control structure and fewer redundancy freedoms than the multi-legged, so the quadruped has become a hot topic in the research field of the robot. The quadruped researchers have been focused on control pattern currently, which can be divided into three categories: control pattern based on model, control pattern based on behavioral and control pattern based on neural network(CPG). But in fact the quadruped is mainly derived from the bionic research of quadruped animal, including bionic structure, bionic gait and bionic control system. It is necessary to build such a kind of control pattern, while the use of bionics research can improve the quadruped more efficiently. The development direction of the quadruped can be summarized as: rapidity, unstructured terrain adaptability and high load. For accomplishing these goals, we introduce the bionics research, and focuses on bionic gait of the quadruped, bionic control system design, Time-Pose control method, gait optimization and event-driven gait control method. These methods can solve motion plan and coordinated control of the quadruped robot. Motion plan consists of swing leg trace plan, gait plan of different kinds of terrain and gait optimization. Coordianted control consists of gait sequence coordination of the torso and single leg, trajectory tracking coordination of the single leg and single joint and cooridination of the quadruped and different kinds of terrain.Bionic gait derives from the bionic research of existing four-legged animal. The difference between the bionic gait and the classic one is the four-legged supportting phase. The advantages of four-legged supporting phase are: 1. improving the stability of quadruped robot, 2. basis of event-driven control method. Also the swing leg planning has been introduced, and proved through theoretical derivation; the bionic study can make the planning reasonable. The swing leg planning has been designed to have the obstacle crossing ability, reduce the maximum joint velocity and joint range of motion.The control system of the quadruped is based on the nervous system of quadruped animal, the control system of the quadruped can be divided into three layers: global planning layer, middle gait planning layer and the bottom executing layer. The three layers equal to brain, middle reflecting layer(CPG rhythm signal) and the bottom layer of the reflecting layer(muscle), respectively. The control system is divided into three layers so as to simplify the control structure, hardware design and the task, and improve the real-time control ability. With the control level reducing, the degree of intelligence declines but the control becomes more precision. The idea comes from intelligent control, it also has been used in our control system design. By dividing the control system intro three layers clearly, we can effectively improve the control effectiveness.The Time-Pose control method has been proposed to manage gait control and pose control, actually corresponds to the middle gait planning layer and the bottom executing layer. A simplified dynamic equation has been used to generate the trajectory of the mass center in gait control layer, and in pose control layer the trajectory is solved to meet the need of the joint trajectory tracking. We assume that the attitude of the quadruped is ideal, so adjustment of the body attitude can be ignored. Since there are redundant degrees of freedom in the forward direction of single leg, redundancy optimization methods are needed. Also we introduce evaluation function for redundancy optimization methods.The event-driven gait control method is expanded from the Time-Pose cotnrol method, environmental sensors and gait policy are added. The core is to achieve adaptability in the unstructured terrain, such kind of adaptability can minimize the impact on the body. And the introduction of the four-legged supporting phase can make the quadruped not change the gait cycle and gait timing, when adapting to the unstructured terrain. The advantage brings a wide range of adaptability in different kinds of terrain.The simulations and experiments show that these mehods are realizable, the feasibility of trotting gait has been tested in RecurDyn simulation software. Also we has completed a variety of gait tests and single leg adaptability experiment based on eventdriven control method. These experiments include flat trotting gait, flat walk, static turning gait, slope static turning gait, trot turning gait, slope walk, obstacle crossing walk and grassland walk. These experiments show the realizability of these methods in this paper. With payload of 50 kg, the quadruped can traveling at the speed of 3.2km/h in trotting gait.