Bipedal walking control by using acceleration factor

Humans can walk efficiently in a certain speed range, while state-of-the-art biped robots can hardly walk as fast as humans. In this paper, we investigate how to walk faster for two simple 2D walking models (Zaytsev in IEEE Trans Robot 34:336-352, 2018), including an inverted pendulum (IP) model and a linear inverted pendulum (LIP) model. Firstly, open-loop analysis is conducted based on the two models. The concept of "acceleration factor" is proposed, which is a key parameter that affects the mid-stance velocity transition between steps. We find that the acceleration factor has a fixed correlation with the velocity transition trend, which is independent of the step length. The step length only affects the feasible range of the acceleration factor, which suggests that we can decouple the control of walking speed from step length. Based on this, we design walking controllers for both models, where the walking velocity and step length are controlled separately. With the proposed controller, closed-loop simulations are performed in the V-REP software, which achieves a mid-stance velocity of 2 and 6 m/s for IP walking and LIP walking, respectively. Besides, walking with varied step lengths along with speed regulation is also demonstrated. This work might be helpful to improve the walking speed of biped robots in the future.