SoftMAC: Differentiable Soft Body Simulation with Forecast-based Contact Model and Two-way Coupling with Articulated Rigid Bodies and Clothes (2312.03297v3)

Abstract: Differentiable physics simulation provides an avenue to tackle previously intractable challenges through gradient-based optimization, thereby greatly improving the efficiency of solving robotics-related problems. To apply differentiable simulation in diverse robotic manipulation scenarios, a key challenge is to integrate various materials in a unified framework. We present SoftMAC, a differentiable simulation framework that couples soft bodies with articulated rigid bodies and clothes. SoftMAC simulates soft bodies with the continuum-mechanics-based Material Point Method (MPM). We provide a novel forecast-based contact model for MPM, which effectively reduces penetration without introducing other artifacts like unnatural rebound. To couple MPM particles with deformable and non-volumetric clothes meshes, we also propose a penetration tracing algorithm that reconstructs the signed distance field in local area. Diverging from previous works, SoftMAC simulates the complete dynamics of each modality and incorporates them into a cohesive system with an explicit and differentiable coupling mechanism. The feature empowers SoftMAC to handle a broader spectrum of interactions, such as soft bodies serving as manipulators and engaging with underactuated systems. We conducted comprehensive experiments to validate the effectiveness and accuracy of the proposed differentiable pipeline in downstream robotic manipulation applications. Supplementary materials and videos are available on our project website at https://damianliumin.github.io/SoftMAC.

• The paper introduces a differentiable simulation framework that reduces penetration anomalies through a forecast-based contact model.
• It leverages MPM along with grid-to-particle anticipation to accurately simulate interactions among soft bodies, rigid bodies, and cloth.
• Experiments validate improved realism in robotic manipulation tasks by mitigating simulation artifacts in diverse material interactions.

Introduction to SoftMAC

In the domain of robotics, simulating physical interactions between various materials is pivotal for training and problem-solving. However, seamlessly integrating materials such as soft bodies, rigid bodies, and cloth in a unified framework presents a noteworthy challenge, particularly when simulating robot manipulation scenarios. A framework named SoftMAC is advancing this field by providing a robust simulation system that handles the intricate dynamics between these different materials.

Technical Innovations

The key to SoftMAC's efficacy is the adoption of the Material Point Method (MPM), which facilitates the simulation of an array of deformable materials. A cornerstone of this framework is the forecast-based contact model, specifically designed to address the contact dynamics for MPM-based simulations. Unlike earlier models, this new contact model incorporates grid-to-particle anticipation to predict possible penetrations and adjusts velocities to mitigate them. This significant reduction in penetration and rebound anomalies results from adjusting the grid velocity based on predicted particle positions, and formulating the contact model as a constrained optimization problem. Additionally, the system employs a penetration tracing algorithm, enabling accurate prognosis and correction of particle penetration in non-volumetric cloth meshes.

Robotic Manipulation Enhancement

SoftMAC's differentiable design is particularly important for robotic manipulation. This differentiation capability ensures that the simulator can provide gradient information across soft bodies, rigid bodies, and cloth interactions, thereby supporting gradient-based optimization strategies. This is valuable for tasks such as trajectory optimization in robotics, where the motion of different mediums needs to be orchestrated to achieve specific goals.

Experimental Verification

Extensive experiments demonstrate the precision and utility of SoftMAC's approach in typical robotic manipulation applications. These include tasks such as pouring liquids, manipulating soft and rigid materials, and cloth handling, often seen in real-world scenarios like cooking or household chores. The paper features performance metrics and qualitative results showing the effective reduction of simulation artifacts like penetration, significantly improving the simulation realism and robustness compared to previous models.

Overall, SoftMAC lays the groundwork for a universal differentiable physics simulation, offering an effective tool for researchers and practitioners in robotics and related fields. The subsequent release of the simulator’s code and documentation will further support advancements in simulation-based robot training and problem-solving.