Contact-Aided Navigation of Flexible Robotic Endoscope Using Deep Reinforcement Learning in Dynamic Stomach

2025-08-29

AI Summary

Flexible robotic endoscopes (FREs) suffer from insufficient navigation stability and accuracy in dynamic gastric environments. Method: This paper proposes a contact-force–feedback–driven deep reinforcement learning (DRL) navigation framework. It innovatively models gastric wall deformability as contact constraints and constructs a high-fidelity dynamic training environment using finite-element simulation. A proximal policy optimization (PPO) algorithm trains the policy network to enable active exploitation of contact forces for robust path planning and target approach. Results: The method achieves 100% target reach rate in both static and dynamic gastric models, with a mean localization error of only 1.6 mm; it maintains an 85% success rate under unseen disturbances—significantly outperforming conventional contact-unaware approaches. This work presents the first closed-loop, contact-driven autonomous navigation framework for FREs, establishing a novel paradigm for precise intervention by soft medical robots.

Technology Category

• Intelligent Robots: Manipulation
• Natural Language Processing: Safety and Robustness
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Navigating a flexible robotic endoscope (FRE) through the gastrointestinal tract is critical for surgical diagnosis and treatment. However, navigation in the dynamic stomach is particularly challenging because the FRE must learn to effectively use contact with the deformable stomach walls to reach target locations. To address this, we introduce a deep reinforcement learning (DRL) based Contact-Aided Navigation (CAN) strategy for FREs, leveraging contact force feedback to enhance motion stability and navigation precision. The training environment is established using a physics-based finite element method (FEM) simulation of a deformable stomach. Trained with the Proximal Policy Optimization (PPO) algorithm, our approach achieves high navigation success rates (within 3 mm error between the FRE's end-effector and target) and significantly outperforms baseline policies. In both static and dynamic stomach environments, the CAN agent achieved a 100% success rate with 1.6 mm average error, and it maintained an 85% success rate in challenging unseen scenarios with stronger external disturbances. These results validate that the DRL-based CAN strategy substantially enhances FRE navigation performance over prior methods.

Problem

• Navigating flexible robotic endoscopes in dynamic stomach environments
• Using contact with deformable walls to reach target locations
• Enhancing motion stability and navigation precision through force feedback

Innovation

• Deep reinforcement learning for flexible endoscope navigation
• Contact force feedback enhances motion stability
• Physics-based FEM simulation for deformable stomach training