Modeling and control of the effects of bilateral asymmetries in insect flight

CMS 273 Project, Winter 2021

Modeling and control of the effects of bilateral asymmetries in insect flight

Michael Dickinson

The Dickinson Lab studies the neural and biomechanical basis of behavior in the fruit fly, Drosophila. We strive to build an integrated model of behavior that incorporates an understanding of morphology, neurobiology, muscle physiology, physics, and ecology. Although our research focuses primarily on flight control, we are interested in how animals transform sensory information into a code that controls motor output and behavior. Recently, we have been developing an entirely computational model of a flying fly, based on 30 years of experiments and observations. We simulate fluid physics and motor control, with the goal of comparing theoretical and empirical kinematics. The model currently uses a classical LTI control system, with principal control modes derived from high-speed video of flight maneuvers. While the classical model is computationally cheap and analytically tractable, it ignores nonlinear and high-frequency dynamics. The fly’s neural control is likely more complex than this classical LTI controller, but perhaps simpler than, say, Model Predictive Control (MPC). The goal of this 273 project is to implement MPC (and maybe other alternative controllers) on the simulated fly and to compare the model's performance to real flies.