Efficient Physically-based Simulation of Soft Bodies in Embodied Environment for Surgical Robot

TL;DR

The paper tackles the lack of open-source, physically-based soft tissue simulation within surgical robot RL platforms. It introduces a MLS-MPM-based soft body simulator with a Neo-Hookean constitutive model, integrated into SurRoL via a modular collision interface and Taichi-based GPU acceleration. The approach achieves interactive rates with physically interpretable tissue dynamics and enables large-scale data generation for policy learning in soft tissue manipulation tasks. This work extends open-source surgical robotics research by enabling deformable tissue simulation in SurRoL and setting the stage for data-driven policy learning in soft-tissue contexts.

Abstract

Surgical robot simulation platform plays a crucial role in enhancing training efficiency and advancing research on robot learning. Much effort have been made by scholars on developing open-sourced surgical robot simulators to facilitate research. We also developed SurRoL formerly, an open-source, da Vinci Research Kit (dVRK) compatible and interactive embodied environment for robot learning. Despite its advancements, the simulation of soft bodies still remained a major challenge within the open-source platforms available for surgical robotics. To this end, we develop an interactive physically based soft body simulation framework and integrate it to SurRoL. Specifically, we utilized a high-performance adaptation of the Material Point Method (MPM) along with the Neo-Hookean model to represent the deformable tissue. Lagrangian particles are used to track the motion and deformation of the soft body throughout the simulation and Eulerian grids are leveraged to discretize space and facilitate the calculation of forces, velocities, and other physical quantities. We also employed an efficient collision detection and handling strategy to simulate the interaction between soft body and rigid tool of the surgical robot. By employing the Taichi programming language, our implementation harnesses parallel computing to boost simulation speed. Experimental results show that our platform is able to simulate soft bodies efficiently with strong physical interpretability and plausible visual effects. These new features in SurRoL enable the efficient simulation of surgical tasks involving soft tissue manipulation and pave the path for further investigation of surgical robot learning. The code will be released in a new branch of SurRoL github repo.

Figures (5)

• Figure 1: The motion sequences of four typical actions for soft body manipulation: pushingmarban2019recurrent, pullingscheikl2022sim, tearingwolper2020anisompm and tensioningou2023sim. All results are simulated and rendered from SurRoL xu2021surrollong2022integratinglong2023humanintheloophuang2023demonstrationguidedhuang2023valueinformed using 30K particles.
• Figure 2: Overview of the System Design. Material point method based soft body simulator is implemented as an extension (blue) of original SurRoL (green)xu2021surrol. Lagrangian particles are generated from mesh or distribution functions before the simulation (red). At each simulation step, the soft body simulator receives the kinematic information of the surgical robot from the physics engine of original SurRoL and performs collision detection. The MPM algorithm receives the collision constraints and performs the simulation, and the results are processed to enhance visual effects. The processed geometry and visual data are then sent back to the PyBullet render to obtain the rendering results (yellow).
• Figure 3: Illustration of Our Collision Detection Method. The top images illustrate the relationship between Eulerian Grids nodes and simulation scene. At each simulation step, the distances from grid nodes to the nearest collision object and the associated object id are recorded into Distance Field and Collision Objects Table respectively. One distance field slice and the corresponding collision objects table slice are displayed at the lower part.
• Figure 4: Results of tissue retraction pore2021learningtagliabue2020soft task under different tissue stiffness. This image displays the simulation results of tissue retraction with four different Young's Modulus (E) values. A higher Young's modulus indicates a stiffer material, while a lower value represents a more flexible or elastic material. All results are simulated using 24K particles and the material density was set to 1000 kg/$m^3$. In the second row of the image, the color represents the degree of deformation. Dark colors indicate the corresponding part of material is squeezed while light colors mean the area is being stretched.
• Figure 5: Performance Statistics. This figure illustrates the simulation time of each frame with various simulation scales and hardware. All simulation results are generated using $64\times64\times64$ grid resolution and each frame consists of 25 sub-steps. The timestep of each sub-step was set to 0.0005s.