Gaussian regressor-based adaptive control of exoskeleton joints in the presence of system uncertainty

System uncertainty remains a challenge for effective control of lower extremity exoskeletons, particularly in clinical populations. Adaptive control offers a potential solution by accounting for unknown system characteristics in real time. Here, we introduce the use of Gaussian-based adaptive control (GBAC) in a two-degree-of-freedom (DOF) exoskeleton for an angular position tracking task in the presence of system uncertainty. The mathematical derivation of the implicitly non-Lyapunov adaptation law is presented using Lagrangian mechanics, including a Gaussian kernel regressor and its stable convergence. We then evaluate GBAC performance in a 2-DOF simulation compared with a previously developed robust adaptive backstepping algorithm, Lyapunov-stable Slotine-Li control, and a proportional-integral-derivative (PID) controller. We additionally complete 1-DOF simulations to evaluate the effects of external disturbance and parameter uncertainty on controller performance. Finally, we evaluate GBAC experimentally in our existing 1-DOF knee exoskeleton along with Slotine-Li and PID controllers. The simulation results demonstrate the improved tracking performance and faster convergence of GBAC, especially in the presence of an external disturbance and uncertainty introduced by extra segment length and mass. The experimental results demonstrate similar performance, wherein GBAC and Slotine-Li provide stable tracking in the presence of unmodeled system dynamics; however, convergence time was faster and tracking error was lower for GBAC. Collectively, these results demonstrate that GBAC is an effective adaptive controller in the presence of system uncertainty and therefore warrants further development and investigation for use in flexible joint exoskeleton systems, particularly those designed for pediatric and/or clinical populations that have inherently high uncertainty.