Postural Control Scheme for Stable Locomotion in Quadruped Robots and Its Optimization

Abstract

What quadruped robots trot stably on various surfaces has still been challenging because they lose balance by the foot bouncing generated by undesirable impulse and the movements of their center of gravity (COG) by change of the surfaces. This paper proposes schemes for stable posture of the quadruped robots, thus enabling quadruped robots to trot on horizontal and slanted surfaces.

For maintaining the postural balance in a region connected by surfaces with different angles, the quadruped robots perform the postural transition scheme (PTS) that regulates the location of the center of gravity (COG) projection point of robots. The PTS is implemented by movement values optimized by a real-coded genetic algorithm (RCGA). The movement values are applied to the centers of the ellipsoidal foot trajectories by using cubic polynomial, which can generate adaptive foot motion continuously.

In addition, for robust posture control, admittance control with impedance modulation (IM) is applied to the foot trajectories, which changes the impedance parameters in real-time depending on the magnitude of the disturbance, such as the excessive swaying of the robot body that causes instabilities for the locomotion. Control thresholds regulated by the angular speed of the robot body are also introduced as a criterion for controlling the excessive swaying. Computer simulations and hardware experiments were performed to verify the effectiveness of the proposed methods.

Meanwhile, this paper also proposes a time-dependent genetic algorithm (TDGA) based on the RCGA to improve the convergence performance of functions over time, such as a foot trajectory. The TDGA has several distinguishing characteristics when compared with the traditional RCGA.

• Individuals are arranged over time, and then the individuals are optimized in sequence. • New search spaces of design variables generate by processes of reductions for the search spaces. • The search spaces for crossover operations are expanded to escape traps of local minima that can occur in new search spaces up to the previous search space before performing any reduction of search space and boundary-mutation operations are performed in the new search spaces.

Computer simulations are performed to verify the convergence performance of the TDGA. The TDGA optimizes the desired foot trajectories of quadruped robots that climb up an uphill trail and the impedance parameters of admittance control of the quadruped robots that trot over irregular terrains. Simulation results evidently present that the convergence performance is improved by the TDGA, which also show that the TDGA could be widely utilized in robot locomotion research.