The Design of an Autonomous Human Tracking Robot with Pacing Applications

Abstract

Given the increasing necessity for modern consumer products to facilitate physical exercise, a robotic pacer was developed to aid runners in their training. The proposed robot uses a multi-sensor approach to identify and track several humans in a cone behind the robot. The motors, chassis, and wheels of a two-wheel-drive robot were first obtained in order to develop and refine the robot's human tracking capabilities. Various sensors were chosen to process human position data using onboard microcontrollers and a Raspberry Pi. A 56 kHz infrared beacon triggers the output of five directional IR modules that continuously determine the human's angle relative to the center of the robot. A facial recognition algorithm was then developed to run on a backwards pointing camera which was connected to the Raspberry Pi. Sensor measurements were combined through use of a sensor fusion algorithm to better estimate the trailing human's position relative to the robot. Lastly, dynamical equations were implemented based on these estimated values to allow the robot to compensate and stay in front of a walking human while the human moved relatively forwards.