Data-Driven Predictive Control for Robust Exoskeleton Locomotion

Abstract

Exoskeleton locomotion must be robust while being adaptive to different users with and without payloads. To address these challenges, this work introduces a data-driven predictive control (DDPC) framework to synthesize walking gaits for lower-body exoskeletons, employing Hankel matrices and a state transition matrix for its data-driven model. The proposed approach leverages DDPC through a multi-layer architecture. At the top layer, DDPC serves as a planner employing Hankel matrices and a state transition matrix to generate a data-driven model that can learn and adapt to varying users and payloads. At the lower layer, our method incorporates inverse kinematics and passivity-based control to map the planned trajectory from DDPC into the full-order states of the lower-body exoskeleton. We validate the effectiveness of this approach through numerical simulations and hardware experiments conducted on the Atalante lower-body exoskeleton with different payloads. Moreover, we conducted a comparative analysis against the model predictive control (MPC) framework based on the reduced-order linear inverted pendulum (LIP) model. Through this comparison, the paper demonstrates that DDPC enables robust bipedal walking at various velocities while accounting for model uncertainties and unknown perturbations.

Figures (9)

• Figure 1: Illustration of data-driven predictive control for bipedal locomotion on lower-body exoskeleton Atalante with various payloads.
• Figure 2: Overview of the proposed layered control framework composed of the DDPC as a planner with constructed data-driven model and low-level controller.
• Figure 3: a) Generalized coordinates for the lower-body exoskeleton Atalante. b) The input and output variables in the x-direction to be used for the Hankel matrix construction.
• Figure 4: One set of the planned CoM and CoP trajectories from the DDPC planner and the tracked trajectory in simulation in right foot frame, the left foot frame trajectories, and the Stance foot. The stance foot frame trajectory in black is generated from DDPC. The corresponding phase variables are plotted in the dashed line.
• Figure 5: Simulation comparison over nominal indicated by blue circles, DDPC indicated by orange stars, and MPC indicated by green diamonds. a) DDPC planner planning trajectories for increasing desired speed, capped at maximum step length $0.2$ m at different step duration $t_d$. b) tracking performance over different desired step length with the same step duration. The dashed line indicate the ideal performance. Error bar indicates the standard deviation over 50 models. c) Simulation time before robot falling for the tracking performance comparison. Error bar indicates the standard deviation over 50 models. The maximum simulation time is $11$ s, indicated by the horizontal dash line. d) Comparison of Nominal and DDPC under time-varying perturbation applied on the negative $x$ direction.