Parameter Identification and Closed Loop Control of a Flywheel Mounted Hovering Robot

A prototype of a hovering multi-terrain mobile robot platform that makes use of a flywheel for stabilization and heading control for rapid maneuverability was developed and presented in a prior paper. It was shown that flywheel stored energy could be transferred to the overall body to generate rapid angular motion once wheel is instantaneously stopped. Solution improved localization accuracy and reduced the overall sensitivity with respect to external disturbances such as non-flat terrain. In this paper, we present a feedback control system to measure dynamic parameters before and after the wheel is stopped. System is designed to follow a predefined path plan and instantaneous torque change causes oscillation after a waypoint is reached. To address this issue, we updated system with an inertial measurement unit (IMU) as a feedback sensor. Then, we investigate the feedback control of individual forward thrust vectors as well as wheel braking timing to minimize amplitude of transient response oscillation and to reduce the steady-state error to an acceptable level that differential drive fans could compensate this error and correct the heading after the rotation around a waypoint occurs. In addition to that, previous mechanical system could transfer all energy stored at once and was not adjustable. In this research, we also investigate varying amount of angular inertia generated by fans and wheel individually and together. To do so, system is modified with stronger forward thrusters. Prior to running the system with a full dynamic model with real mechanism, we implemented a simulation to empirically extract system parameters and adjust controller gains to follow a predefined path with open and closed loop control schemas with objective of minimizing localization error. Finally system is tested with real mechanism. Governing equations, simulation and empirical results comparison are presented and generated trajectories of various simulation and real world settings are listed. Test results verify that, with a closed loop control system, overshoot and total error about a waypoint can be minimized to an acceptable level at and after transient response phase.