A compliant humanoid walking strategy based on the switching of state feedback gravity compensation controllers

Abstract

This paper provides stability analyses for two different types of desired gravity compensation controllers, employing both motor and link feedback, and describes a means by which these controllers can be used to control a compliant humanoid robot in order to ensure the successful execution of walking trajectories. Given the challenging task of controlling compliant bipedal systems, owing to their possession of under-actuated degrees of freedom, the full actuator and link dynamics are accounted for. The proposed walking strategy involves a process of switching between three distinct controllers which is contingent upon the force feedback provided by the force/torque sensors embedded in the robot's feet. These controllers were tuned using a simulation model of the robot and were then implemented on the compliant COMAN legs, whose performance of walking confirms the controllers' stability, in addition to the walking scheme's efficacy.

Lyapunov Stability Margins for humanoid robot balancing

“…Using control law (5), the closed-loop system was rendered globally asymptotically stable, as was demonstrated in the analysis presented in [23]. Hence, the following Lyapunov function shall be considered throughout the paper:…”

“…where has been replaced with since we are considering a desired gravity compensation controller, while ( ) + denotes the Moore-Penrose pseudoinverse of . By considering the closed-loop system's steady state equations, we can define the following matrix that plays a central role in the stability analysis since it must satisfy certain conditions in order for closed-loop stability to be guaranteed, as shown in [23]:…”

“…Explicit calculations proving that the time derivative is negative definite are omitted here as they can be found in [22]. It is essential to state that global stability holds if and only if the following conditions are satisfied [23]:…”

Selective-Compliance-Based Lagrange Model and Multilevel Noncollocated Feedback Control of a Humanoid Robot

“…A method by which the GRFs can be translated into joint torques is described by [26], mathematically proving this specific controller's passivity. The method proposed in this paper is an extension to the work seen in [9], which involves the use of PDD control (motor position, motor velocity and link velocity feedback) on 6 degree-offreedom (DOF) SS and 3-DOF DS models of the robot, in combination with gravity compensation control. The mathematical models describe the full compliant dynamics (the motor and link dynamics appearing before and after the elastic element), thereby accounting for the mere sagittal elements, since it is only those joints that comprise Series Elastic Actuators (SEA).…”

Dynamically transitioning between surfaces of varying inclinations to achieve uneven-terrain walking

“…Its intrinsic compliance, due to the passive elasticity in its joints, makes this robot a good candidate for this kind of task. In previous works [6], [7], [8] we focused on the effects of compliance on the dynamics of walking. It was demonstrated that compliance can reduce the effect of disturbances due to the take off and landing of the foot.…”