Development of a colon endoscope robot that adjusts its locomotion through the use of reinforcement learning.

PURPOSE: Fibre optic colonoscopy is usually performed with manual introduction and advancement of the endoscope, but there is potential for a robot capable of locomoting autonomously from the rectum to the caecum. A prototype robot was designed and tested.
METHODS: The robot colonic endoscope consists in a front body with clockwise helical fin and a rear body with anticlockwise one, both connected via a DC motor. Input voltage is adjusted automatically by the robot, through the use of reinforcement learning, determining speed and direction (forward or backward).
RESULTS: Experiments were performed both in-vitro and in-vivo, showing the feasibility of the robot. The device is capable of moving in a slippery environment, and reinforcement learning algorithms such as Q-learning and SARSA can obtain better results than simply applying full tension to the robot.
CONCLUSIONS: This self-propelled robotic endoscope has potential as an alternative to current fibre optic colonoscopy examination methods, especially with the addition of new sensors under development.