Real-Time Motion Tracking System and Data Analytics for Live Insects Using Three-Wheeled Servosphere System

Abstract

Tracking the motion and pathing of insects is critical for understanding the underlying factors determining their behaviors. Methods of tetherless motion tracking using servosphere systems as omnidirectional treadmills have been demonstrated to be as suitable for insect tracking as conventional methods, such as tethers or markers in arenas or trackballs, while presenting several practical and experimental advantages. This project expands on previously established applications of the servosphere system as a method of motion tracking with three primary contributions: (i) building, coding, and operating a three-wheeled servosphere system in order to evaluate its viability as a method of live motion tracking, (ii) tracking the motion of subjects stimulated by external olfactory cues to show the system's robustness with both innate and stimulated motion, and (iii) developing a convolutional neural network (CNN) model using trajectories recorded by the servosphere system to demonstrate its potential applications in classification via artificial intelligence. Across seven different subjects, the average error value was 3.23 mm, which fell within the allotted uncertainty of 6 mm. The unstimulated subjects' average speed was calculated to be 5.448 mm/s and the stimulated subjects' average speed was 11.594 mm/s, and their trajectories yielded a model accuracy of 64%. The experimental results show promising signs of applying the servosphere system in further motion tracking data analytics.