Vision-Language Procedural Reasoning for Context-Aware Reward Modeling of Robotic Endovascular Guidewire Navigation

2026-06-29

AI Summary

This work addresses the limitations of existing robotic guidewire navigation systems that rely on static reward functions and struggle to adapt to dynamic task demands in complex vascular environments. To overcome this, the authors propose a Vision-Language Process Reasoning (VL-PR) framework, which introduces multimodal large language models into endovascular navigation for the first time. By interpreting real-time visual inputs, VL-PR infers high-level anatomical context and procedural stage, enabling dynamic adjustment of reward function weights and facilitating context-aware reinforcement learning. Evaluated on a physical robotic platform, the approach significantly improves both success rate and navigation efficiency compared to conventional static reward strategies, offering a scalable and globally consistent solution for autonomous, multi-task vascular interventions.

Abstract

Robotic-assisted endovascular interventions demand accurate, stable, and context-aware guidewire navigation in complex and patient-specific vascular anatomies. Despite recent advances in robotic precision and learning-based control, existing autonomous navigation methods remain limited by their reliance on static reward functions and the lack of explicit procedural reasoning regarding anatomical context and task progression. To address these challenges, this paper proposes a vision-language procedural reasoning (VL-PR) framework for autonomous guidewire navigation. The framework integrates a multimodal large language model (MLLM) as a procedural reasoning module that interprets real-time visual observations to infer high-level navigation contexts. Instead of generating low-level control commands, the inferred procedural insights enable context-aware reward adaptation by dynamically adjusting the importance of reward components across different navigation phases. This approach allows a single policy to resolve competing objectives and handle complex transitions while preserving a consistent global task goal. Experiments on a physical robotic platform across diverse vascular scenarios demonstrate enhanced task reliability and streamlined navigational efficiency, highlighting the advantages over static-reward methods and offering a scalable solution for complex and multi-task robotic endovascular procedures.

Problem

• vision-language procedural reasoning
• context-aware reward modeling
• robotic endovascular navigation
• autonomous guidewire navigation
• multimodal reasoning

Innovation

• vision-language procedural reasoning
• context-aware reward modeling
• multimodal large language model
• autonomous guidewire navigation
• dynamic reward adaptation