Chuzhao Liu

Stability analysis of quadruped robot based on compliant control

Stability is a key point in developing quadruped robot. In this paper, we focus on allowing heavy quadruped robot to be self-adaptive to various disturbances. The Spring Loaded Inverted Pendulum method is applied in regulating feet landing positions. Moreover, yaw motion is discussed to revise direction-deflection after disturbance of external force. In dynamics analysis, both inverse dynamics and balance controller are employed to predict desired torque of joints. Meanwhile gains of position controller can be decreased. The main contribution is an active compliant controller with respect to robot posture. According to linear functions, it can adjust four leg lengths separately to keep robot balance. Therefore, stability is kept both in torque level and position level. The simulation results prove the effect of proposed control scheme under disturbances of external force and uneven terrain. In summary, this control scheme is exceedingly useful in enhancing legged robot stability.

Real-time projection and dynamics analysis of quadruped robot

Mobility and stability of quadruped robot need to be explored. A novel real-time leg trajectory projection method is proposed based on Spring-Loaded Inverted Pendulum and virtual leg algorithm. Instead of only planning landing position, optimized Spring-Loaded Inverted Pendulum method employs three simple factors, touchdown angle, liftoff angle and foot landing position, to control velocity of robot with multi-joint leg. Without extra torque, swing motion of torso is determined by current status of robot passively. Whole body dynamics are discussed. Feedback, feedforward and balance controllers are employed to generate desired torque. Trotting simulation shows the proposed methods are sufficient to achieve trotting at a speed of 1m/s. Experiment at the same speed indicates a desirable mobility and stability. Results help us have a better understanding about the principle of quadruped mammals moving. The proposed method is an expansion of virtual leg algorithm, which can be applied to various quadruped robots with bionic structure.