Refined Motion Compensation with Soft Laser Manipulators using Data-Driven Surrogate Models

2024-07-02

arXiv.org

AI Summary

To address targeting inaccuracies in hepatic tumor laser ablation caused by respiratory and cardiac-induced organ motion, this paper proposes a data-driven surrogate-model-based motion compensation method for soft robotic manipulators, enabling non-contact, precise thermal ablation. The method innovatively integrates spectral submanifold (SSM) theory with model predictive control (MPC) to establish a low-dimensional, transferable dynamic modeling framework for soft actuators—supporting plug-and-play compatibility across diverse configurations. By synergistically combining SSM-based dimensionality reduction, surrogate modeling, and MPC optimization, the approach achieves significantly improved positioning accuracy over baseline methods in simulation, satisfies real-time closed-loop control requirements, and demonstrates seamless interchangeability across multiple soft robotic manipulator designs.

Technology Category

• Intelligent Robots: Manipulation
• Machine Learning: Learning with Manifolds
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Non-contact laser ablation, a precise thermal technique, simultaneously cuts and coagulates tissue without the insertion errors associated with rigid needles. Human organ motions, such as those in the liver, exhibit rhythmic components influenced by respiratory and cardiac cycles, making effective laser energy delivery to target lesions while compensating for tumor motion crucial. This research introduces a data-driven method to derive surrogate models of a soft manipulator. These low-dimensional models offer computational efficiency when integrated into the Model Predictive Control (MPC) framework, while still capturing the manipulator's dynamics with and without control input. Spectral Submanifolds (SSM) theory models the manipulator's autonomous dynamics, acknowledging its tendency to reach equilibrium when external forces are removed. Preliminary results show that the MPC controller using the surrogate model outperforms two other models within the same MPC framework. The data-driven MPC controller also supports a design-agnostic feature, allowing the interchangeability of different soft manipulators within the laser ablation surgery robot system.

Problem

• Laser Ablation
• Respiratory and Cardiac Motion Compensation
• Tumor Tracking

Innovation

• Spectral Submanifold Theory
• Non-contact Laser Ablation
• Soft Robotic Arm Control