Motion Compensator for an Untethered Walking Insect Using Adaptive Model Predictive Control

Abstract

A locomotion compensator is normally utilized to observe the behavior of walking insects. These compensators cancel out the movement of freely walking insects to facilitate long-term imaging for studying behavior. However, controlling the locomotion compensator with a small error (≤ 1 mm) has been challenging due to the random motion of walking insects. This study introduces an adaptive model predictive control (MPC) approach combined with trajectory prediction to effectively control the transparent omnidirectional locomotion compensator (TOLC) for a randomly walking fire ant. The proposed MPC with prediction (MPCwP) utilizes the average velocity from the previous gaiting cycle to estimate its future trajectory. Experimental results demonstrate that MPCwP significantly outperforms MPC without prediction (MPCwoP), which relies solely on the current position and orientation. The distance error of the MPCwP method remains below 0.6 mm for 90.3% and 1.0 mm for 99.2% of the time, whereas MPCwoP achieves this only 32.6% and 69.1% of the time, respectively. Furthermore, the proposed method enhances the tracking performance of the heading angle, with the heading angle error staying below 8 deg for 92.6% of the time (wθ=1.0). The enhanced performance of the proposed MPC has the potential to improve the observation images and enable the integration of additional equipment such as an optical microscope for brain or organ imaging.