Adaptive cable-driven parallel load simulator inspired by physical reservoir computing for folding wings

Abstract

Predicting the effects of aerodynamic loads using equivalent wrenches generated by a load simulator is essential for the expedited development of folding wings. However, large-angle rapid deployment characteristics of folding wings hinder the direct application of existing load simulators. Inspired by physical reservoir computing, we propose a novel physical computing network capable of adaptively simulating dynamic torque and radial forces while suppressing surplus torque. This approach facilitates the development of a reconfigurable load simulator design for folding wings. The network is established based on an analysis of the nonlinear loading characteristics of a cable-driven parallel robot limb actuated by elastic element, and its simulation capabilities are evaluated. An optimization design method is proposed, significantly reducing the required number of limbs through parameter adjustments. The simulation capabilities of the load simulator designed using this method are evaluated through numerical simulations, and an experimental load simulator is constructed for loading tests. Results from both multi-condition simulations and experiments demonstrate that the load simulator achieves high accuracy across various load scenarios, with an R-squared value exceeding 0.99.