Modeling and LQR control of insect sized flapping wing robot

Abstract

Flying insects perform agile maneuvers like backflips, sharp turns, and collision recovery. Replicating these in sub-gram flying insect robots (FIRs) requires fast and responsive control systems. Current FIRs rely on elaborations of proportional-integral-derivative (PID)-type feedback control, requiring painstaking tuning and task-specific adjustments for complex maneuvers. Optimal control offers a solution but has been limited by approximate models or computational demands unsuitable for onboard implementation. Here, we used a more accurate stroke-averaged model of forces and torques, derived from a sensitive two-axis torque sensor, to implement the first demonstration of optimal control on an FIR that is computationally efficient enough to be performed by a microprocessor carried onboard. Applied to the 150 mg UW RoboFly, this enabled stable hovering (RMS error 2.5 cm) and trajectory tracking at speeds up to 25 cm/s using a linear quadratic regulator (LQR). These results were enabled by a more accurate model and laid the groundwork for integrating low-power receding-horizon control to achieve aggressive maneuvers in FIRs.