Model-Based Control Allocation During State Transitions of a Variable Recruitment Fluidic Artificial Muscle Bundle

A model-based control scheme for state transitions of a variable recruitment fluidic artificial muscle (FAM) bundle is developed and experimentally validated. FAMs can be bundled together in parallel to exhibit variable recruitment functionality, which is an activation strategy inspired by how motor units (MUs) in skeletal muscle are recruited. By adapting variable recruitment, an FAM bundle is able to operate efficiently over its entire force-contraction space while increasing control authority and bandwidth at low recruitment states. A variable recruitment bundle poses a hybrid control problem as it operates by controlling pressure as a continuous variable while simultaneously shifting between discrete recruitment states. During such state transitions, the bundle may experience a lag in strain if the shift timing is not properly anticipated. In this study, a model that captures the interaction effects between FAMs and a hydraulic system model is used to inform the controller of when a state transition should be made. The proposed control scheme is compared to a baseline control scheme that uses a percentage of the source pressure as the threshold for when a shift is made. The controller performance is evaluated by tracking a sinusoidal strain trajectory, and the average and maximum strain errors are compared between the baseline and proposed controller. The applied FAM pressures are presented to show that the model-based compensation is able to determine when a transition needs to be made. As a result, the tracking performance of the proposed control scheme is shown to significantly decrease the integrated absolute and maximum errors.