Navigation Through Endoluminal Channels Using Q-Learning

TL;DR

This paper introduces a Q-learning based method for autonomous navigation within bronchial tubes, demonstrating effective strategy learning in simulation to aid medical robotic applications.

Contribution

It presents a novel reinforcement learning approach for endoluminal navigation, including environment simulation and experimental validation for bronchial tube navigation.

Findings

Agent successfully learned navigation strategies

Achieved goal-reaching in simulated environment

Potential for real-world medical robotic use

Abstract

In this paper, we present a novel approach to navigating endoluminal channels, specifically within the bronchial tubes, using Q-learning, a reinforcement learning algorithm. The proposed method involves training a Q-learning agent to navigate a simulated environment resembling bronchial tubes, with the ultimate goal of enabling the navigation of real bronchial tubes. We discuss the formulation of the problem, the simulation environment, the Q-learning algorithm, and the results of our experiments. Our results demonstrate the agent's ability to learn effective navigation strategies and reach predetermined goals within the simulated environment. This research contributes to the development of autonomous robotic systems for medical applications, particularly in challenging anatomical environments.