Real‐time and precise insect flight control system based on virtual reality

Abstract

A non‐invasive and automatic control method based on real‐time virtual reality and reinforcement learning is reported to modulate insects’ flight behaviour. The online configurable LED display system is proposed to present the visual stimulus within ultra‐low latency. State‐action‐reward‐state‐action (SARSA) coupled with sequential K ‐means clustering is applied to generate optimal control sequence automatically to ensure the satisfactory control performance. By evaluating the control results of 15 bumblebees, it shows that the proposed method achieves to modulate the abdominal waving pattern quickly (less than six steps) with small deviation (<0.03).