Real-Time Planning and Nonlinear Control for Quadrupedal Locomotion With Articulated Tails

Abstract

The primary goal of this paper is to develop a formal foundation to design nonlinear feedback control algorithms that intrinsically couple legged robots with bio-inspired tails for robust locomotion in the presence of external disturbances. We present a hierarchical control scheme in which a high-level and real-time path planner, based on an event-based model predictive control (MPC), computes the optimal motion of the center of mass (COM) and tail trajectories. The MPC framework is developed for an innovative reduced-order linear inverted pendulum (LIP) model that is augmented with the tail dynamics. At the lower level of the control scheme, a nonlinear controller is implemented through the use of quadratic programming (QP) and virtual constraints to force the full-order dynamical model to track the prescribed optimal trajectories of the COM and tail while maintaining feasible ground reaction forces at the leg ends. The potential of the analytical results is numerically verified on a full-order simulation model of a quadrupedal robot augmented with a tail with a total of 20 degrees of freedom. The numerical studies demonstrate that the proposed control scheme coupled with the tail dynamics can significantly reduce the effect of external disturbances during quadrupedal locomotion.

Distributed Planning of Collaborative Locomotion: A Physics-Based and Data-Driven Approach

“…One downside to this approach is that it generally necessitates quasi-static locomotion due to the zero moment point (ZMP) criterion [27] placed on the center of pressure (COP). Furthermore, in the context of collaborative locomotion, a major pitfall is that this template model cannot accommodate torques caused by external forces induced about the COM [28], [29], which is important when planning for cooperative dynamic gaits of holonomically constrained multi-agent systems.…”

Layered Control for Cooperative Locomotion of Two Quadrupedal Robots: Centralized and Distributed Approaches

“…In contrast to full-order models of legged locomotion, template models present reduced-order representations of legged robots that significantly reduce the computational burden and complexity associated with trajectory optimization. Various template models, including LIP [41], SRB [43]- [45], and centroidal dynamics [42], have been successfully integrated with the MPC framework for the real-time planning of bipedal and quadrupedal robots [43]- [53], [55]. The main challenge with using template models is bridging the gap between reduced-and full-order models of locomotion arising from abstraction (e.g., ignoring the legs' dynamics in template models).…”

Distributed Data-Driven Predictive Control for Multi-Agent Collaborative Legged Locomotion

“…Nonetheless, the model's simplicity has been of great interest over the years and has been used extensively in both simulation and experiments using various platforms [12]- [16]. It is also worth mentioning that this model is not amenable to adding moments induced about the center of mass (COM) by external forces [17], [18], which is important when dealing with holonomically constrained multi-agent systems.…”