基于虚拟黏弹性肌肉模型的仿人机器人不平整地面稳定行走

In complex unknown environments, it is difficult for humanoid robots to obtain exact terrain information during walking, which leads to the errors between the planned footholds and the actual foot landing positions. It brings great disturbance to the balance of robots. To address this problem, a virtual muscle model imitating the viscoelasticity of human muscle is proposed, and a stable walking control method for humanoid robots on uneven terrain is designed. Firstly, from the perspective of bionics, a virtual muscle model with ability of extension and shrink is established by extending the traditional muscle model, and its viscoelasticity is analyzed. Then, a muscle length and force controller is designed with LQR (linear quadratic regulator) method based on this model. Finally, combining the feedback information of force sensors on the feet, the proposed model is applied to adjusting the height of feet during walking for humanoid robots, which guarantees robots to adapt to the abrupt change of terrain height in complex unknown environments, or to accomplish stable walking when the terrain information obtained by the sensor system seriously differs from the actual condition. The result shows that the stable walking of humanoid robots can be realized on an uneven terrain within height errors of 6 cm and at speed of 1.8 km/h via the proposed method. Its effectiveness is validated by walking simulation on BH6-6P platform.