On The Dynamic Walking Theory And System Based On Bionics And Passive Walking

The walker with low energy consumption has great social significance and application value. Passive walking which proposed by the McGeer is a different way from the traditional research ideas. It emphasizes the study of essential characteristics of human walking. Dynamic walking theory based on passive walking is a viable option to achieve low-power.The theoretical research on passive dynamic walking has begun systematic. However, there are still many issues which need researchers in this direction to face together. How to improve the response to rough road robot immunity under the premise of low-power? How to explore the mechanism of human walking and to design the control method which is easier to implement for the prototype with bionics? How to make robots walk with different modes like humans?This paper aims to study dynamic walking theory which is easy to be applied in real-time simulation platform or physical system with passive walking and bionics. Creating a point mass walking model has significance to theoretical research of passive dynamic walking. Compass-like model plays an important role to promote the early development of passive dynamic walking. However, the compass-like model cannot make the correct response to the effect of the human ankles. And it is not reasonable to add the larger torque between the leg and the ground in some studies. The compass-like walker with controllable flat feet has been proposed and modeled. It is composed by the compass-like model and two rotatable foot rods. The best energy ratio of leg length and foot length has been obtained by the simulation results. The best energy ratio is 0.18 which matches the ratio of human and it can be used to guide the selection of the foot length in the following theoretical studies and the design of physical prototypes.So far, most of the control methods for passive dynamic walker are efficient, but not easy to realize in the physical machine as the output torque is complex. Minimalistic torque function is proposed through observation and analysis of pushing the swing. The function should satisfy conditions: function values does not change sign in the definition domain, constant positive or constant negative (including zero); there can be up to one continuous non-zero function values in an action period. By this function structure, the input torque can do positive work only and the frequency of the actuator can be low. Minimalistic torque function with feed-forward can make compass-like with feet model and"PADW2"physical prototype walk efficiency and stability in some constant road conditions, but disturbance rejection is poor. The control torque function for the whole movement is planned before the action by human. And the road environment should be known in advance in efficient walking. These can be seen in common lives. Accordingly, NPFM (Neural network Predictive control and Feed-forward switching based on Minimalistic torque functions) is proposed. This control method can significantly improve the robot's immunity to environmental changes and can make compass-like with feet model walk efficiency and stability in constant road conditions and in random bumps. In addition, this control method has low training complexity and simple torque functions, which are easy for controller design and actuator implementation on physical robot.In the process of human walking, moving back and forth of upper body plays a key role to maintain balance and to change speed. Research the abstract model with rotational upper body could help to understand the nature of human walking and could be essential basic theory to develop new robot. Model with rotational upper body is composed by the Compass-like model and a rod connected in the hip joint. Since the addition of rotational upper body, the walker has no stable passive gait. Wheel-like method can make the model with rotational upper body walk stably, which is proposed according to actual human walking. Just make sure the swing leg can touch the ground and the two legs have the same length when the leg swing over a certain angle, the two-dimensional biped walker can walk stably under the limited speed. With minimalistic torque function, control torque of wheel-like method can be simple and low energy consumption.With minimalistic torque function and neural network predictive control, wheel-like method can make the model with rotational upper body walk efficiency in some typical conditions. However, impact of rotational upper body has not been considered. Rimless wheel model with upper body is considered to study the pure impact of the upper body. There is only one torque input added between the upper body and the wheel. The upper body is limited to the nearby of the equilibrium position by spring -damping torque. Equilibrium position ??0 can affect the model's moving speed and the higher speed needs the higher energy consumption. Model will eventually stop whenα0≤12°. Speed will convergence to a non 0 value when 13°≤α0≤15°. Speed will never convergence whenα0≥16°. Optimization of equilibrium position can reduce the torque of the support ankle.Wheel-like method can make the virtual prototype walk efficiency on the level with the angle limit of mechanical structure and the reflective torque control. The average speed is up to 0.62m/s and the energy consumption ratio of moving is about 0.12.