Model reference adaptive control of experimentally identified honeybee visual tracking dynamics

Abstract

Insects navigating environments containing visual targets are used as model systems for computationally constrained micro aerial vehicles due to their visually-mediated dynamic responses to those targets, which may include adaptive and time-varying controller structures. In this study, honeybees were induced to track a vertically moving target (hive entrance). To measure the trajectories, a three-camera real-time automatic tracking system was used to track and record multiple insects' three-dimensional positions and velocities. Frequency domain system identification was then used to identify the associated dynamic systems from the 3D position information. These results indicate the trajectories are described by dynamic system variation coupled with visuomotor delays distributed from 2-100 ms. The identified transfer function model structures include those with potentially challenging control design problems, including delays, non-minimum phase systems, and unstable dynamics. A model-based adaptive controller is implemented on unstable and non-minimum phase systems. The results for three different reference models show that the unstable system can follow the desired trajectories. This study serves as a structure for studying honeybee trajectory control and its relationship to adaptive control in a repeatable environment.