Unisoma: A Unified Transformer-based Solver for Multi-Solid Systems

Shilong Tao, Zhe Feng, Haonan Sun, Zhanxing Zhu, Yunhuai Liu

TL;DR: We propose Unisoma to solve multi-solid systems in an explicit modeing paradigm.

Abstract: Multi-solid systems are foundational to a wide range of real-world applications, yet modeling their complex interactions remains challenging. Existing deep learning methods predominantly rely on implicit modeling, where the factors influencing solid deformation are not explicitly represented but are instead indirectly learned. However, as the number of solids increases, these methods struggle to accurately capture intricate physical interactions. In this paper, we introduce a novel explicit modeling paradigm that incorporates factors influencing solid deformation through structured modules. Specifically, we present Unisoma, a unified and flexible Transformer-based model capable of handling variable numbers of solids. Unisoma directly captures physical interactions using contact modules and adaptive interaction allocation mechanism, and learns the deformation through a triplet relationship. Compared to implicit modeling techniques, explicit modeling is more well-suited for multi-solid systems with diverse coupling patterns, as it enables detailed treatment of each solid while preventing information blending and confusion. Experimentally, Unisoma achieves consistent state-of-the-art performance across seven well-established datasets and two complex multi-solid tasks. Code is avaiable at [https://github.com/therontau0054/Unisoma](https://github.com/therontau0054/Unisoma).

Lay Summary:

We want to understand how multiple solid objects, like metal parts and flexible materials, interact and deform when they come into contact. This is important for many real-world applications, such as robotic gripping, metal stamping, or even medical simulations. However, existing AI (artificial intellengence) models often struggle to handle the complex interactions when many different solids are involved.

To tackle this, we build a new system called Unisoma. Instead of letting the AI learn everything by trial and error (which can lead to confusion when too many objects are involved), we teach it to explicitly recognize and model key physical factors — like objects touch others and forces are applied. Our method uses a powerful type of foundational model called a Transformer to manage all this in a flexible way that works with different numbers and types of objects.

We find that Unisoma is not only more accurate than existing methods, but also more efficient and better at handling unseen scenarios. This makes it a promising tool for engineers and scientists who want to simulate physical systems more reliably. We've made the code freely available to help others apply this approach to their own challenges.

Application-Driven Machine Learning: This submission is on Application-Driven Machine Learning.

Link To Code: https://github.com/therontau0054/Unisoma

Primary Area: Applications->Chemistry, Physics, and Earth Sciences

Keywords: Deep Learning, Multi-solid Systems, Explicit modeling

Submission Number: 4840