Provably Robust Learning-Based Approach for High-Accuracy Tracking Control of Lagrangian Systems
share
Inverse dynamics control and feedforward linearization techniques are typically used to convert the complex nonlinear dynamics of robots to a set of decoupled double integrators, and then a standard, outer-loop controller can be used to calculate the commanded acceleration for the linearized system. However, these methods typically depend on having a very accurate system model, which is not available in practice. While this challenge has been addressed in the literature using different learning approaches, most of these approaches do not provide safety guarantees in terms of stability of the learning-based control system. In this paper, we provide a novel, learning-based control approach based on Gaussian processes (GPs) that ensures both stability of the closed-loop system and high-accuracy tracking. In our proposed approach, we use GPs to approximate the error between the commanded acceleration and the actual acceleration of the system, and then use the predicted mean and variance of the GP to calculate an upper bound on the uncertainty of the linearization. This uncertainty bound is then used in a robust, outer-loop controller to ensure stability of the overall system. Moreover, we show that using our proposed strategy, the tracking error converges to a ball with a radius that can be made arbitrarily small through appropriate control design. Furthermore, we verify the effectiveness of the proposed approach via simulations on (i) a 2 degree-of-freedom (DOF) planar manipulator, and (ii) a 6 DOF industrial manipulator using the Gazebo simulator.
READ FULL TEXT