Nonholonomic Yaw Control of an Underactuated Flying Robot With Model-Based Reinforcement Learning

Nonholonomic control is a candidate to nonlinear systems with path-dependant states. We investigate an underactuated flying micro-aerial-vehicle, the ionocraft, that requires nonholonomic in yaw-direction for complete attitude control. Deploying analytical law involves substantial engineering design and sensitive inaccuracy system model. With specific assumptions on assembly dynamics, we derive Lie bracket yaw of ionocraft. As comparison significant effort required analytic law, implement data-driven model-based reinforcement learning controller simulated flight task. demonstrate simple framework can match derived - rate chosen actions few minutes data, without pre-defined dynamics function. This letter shows learning-based approaches are useful as tool synthesis laws previously only addressable through expert-based design.