On the Dynamic Walking Theory and System Based on Bionics and Passive Walking
|Course||Control Theory and Control Engineering|
|Keywords||Walking robot Passive dynamic walking Bionics Simple torque function Upper equilibrium position Imitation round walking Energy consumption|
The walking mechanism of the low-power operation of great social significance and application value. McGeer proposed based on passive principle of dynamic walking is a different from the traditional active bipedal walking research ideas more emphasis to the essential characteristics of the study of human walking. Based on passive dynamic walking walking principle is to achieve a workable solution for the low power nature walk. Multinational researchers study has begun to take a systematic, dynamic walking based on the principle of passive walking However, there are still many issues that face researchers in this direction. (1) in the maintenance of low energy, further improve the stability of the walk as well as deal with the rough road immunity; ② human walking the internal mechanism further explore the use of bionics design easier to achieve control method and the robot prototype; ③ so The robot like humans have a variety of walking mode. This paper aims to in-depth system analysis based on the mechanism and function of human walking system to achieve the use of passive walking the the dynamic walk theory principles and bionics research easy application of highly efficient and stable system in real-time simulation platform or bipedal walking robot walking. Abstract point mass model based on dynamic walk theory based on the principle of passive walking significance, the compass robot point foot model promotion of played an important role in the early development of the walking machine based on the principle of passive dynamic walking. Compass robot point-foot model and not the right response to the role of the human ankle, the leg link with the ground in some studies to larger torque is unreasonable. So with the compass robot controllable foot rod abstract model is proposed and established, which consists of a compass robot and two rotating foot rod. By changing the robot feet long and driving energy range search can make the robot walk stably ankle, foot length and leg length to get the best energy ratio of about 0.18. This ratio with the measured ratio of the human to match, it can be used to guide the subsequent theoretical research and a prototype design of the entity's foot rod length selection. So far, most of the walking machine control method based on the principle of passive dynamic walking are more efficient, but the output torque changes complex and difficult to achieve on the physical machine. Based on the observation and analysis of the human push swing action, simple torque function is proposed, this function should satisfy the conditions: the value of the function the same number in the domain is defined, constant positive or constant negative (including zero); period of an action, only a maximum of period of continuous non-zero function value. Construct this driver function format, the torque input only do positive work, no energy is wasted; low frequency requirements of the actuator, simple. Efficient and stable walking function format of the drive to the simple feedforward applied to the foot the compasses robot model with PADW2 entities prototype allows the walking machine unchanged within a certain range of road conditions, but the immunity is weak. Common example through the analysis of two life: put empty buckets and eyes closed, walk, come ① human action before planning control moment function of the entire action; advance awareness ② efficient human walking road environment. As a result, based on a simple moment function neural network predictive control method is proposed. This control method can significantly improve the immunity of the robot on the road to environmental change, natural and constant road conditions, changes in road conditions, as well as random bumps on the road and efficient stable walking the compass robot model with legs. In addition, this control method algorithm is simple, the difficulty of training and has a simple function of the torque, easy-to-physical robot controller design and the realization of the actuator. Upper body center of gravity moves back and forth in the process of human walking, keeping balance and change the walking speed played a key role. The activities controlled study with the abstract model of the upper body helps to understand the nature of human walking, the basic theory is indispensable to the development of new robots in the future. Abstract walking with activities in the upper body model by compasses robot model, as well as a hinge in the connecting rod of the hip. Due to the activities of the upper body join, not pure passive gait of walking machine. Abstract model of imitation round walking strategy can make the link with activities in the upper body stable walking, it is based on actual human walking legs to imitate wheels: limited speed, as long as the supporting leg placed certain angle, swing leg effective the ground, and the swing leg and the supporting leg length is the same, the two-dimensional bipedal walking machine can walk stably. Application imitation wheel walking strategy, given to different target swing angle, can produce different step, but at the same time need to target speed given there are corresponding changes. Combined with a simple torque function, imitation round walking strategy to control torque and lower energy consumption. Imitation round walking strategy combined with a simple torque function and nonlinear model predictive control abstract model can make a link with activities in the upper body stable while walking in some of the typical operating conditions, but did not consider the upper rod to the walking machine dynamic brought impact. To study pure upper body on walking, with the endless round of activities upper body model is considered. The model has only one acting on the upper rod and boundless wheel torque control input. Combined to calculate the compensation in the form of a spring damping torque of the upper rod limit near the equilibrium position. Change the equilibrium position of the upper body can affect the traveling speed of the endless wheel model of the upper body with activities. Α0 ≤ 12 ° in the equilibrium position, the model will eventually stop. In 13 ° ≤ α0 ≤ 15 °, the speed of convergence in a non-zero value. α0 ≥ 16 °, the speed of convergence, that is unstable. But can be used for a short time when the actual walking acceleration. Optimize the activities of the equilibrium position of the upper body can also reduce the abstract model with the activity upper body connecting rod supporting ankle torque. The virtual prototype model of imitation round walking strategy combined mechanical structure limit angle, torque control, reflected by similar conditions with movable upper rod and the plains stable, efficient walking. Changing the upper rod balancing location can change the traveling speed of the virtual prototype, the higher speed requires a larger energy consumption support. Typical average speed of 0.62m / s can be reached about moving the energy consumption rate of about 0.12.