π A New Paper is on the Way to a Top Conference!
π Exciting news! A new paper is about to be submitted to a top-tier conference!
In this work, we propose an improved modeling framework for heterogeneous hybrid objects, based on advancements to the Spring-Gaus physics engine. The goal is to enable more accurate and physically consistent simulation of real-world object interactions in embodied AI scenarios.
In this article, I was responsible for constructing Unity-based physics simulation data, modifying the graph neural network (GNN) for the physics engine, conducting experimental validation, and contributing to the paper writing. Additionally, I participated in the creation of the first dataset for embodied physics simulation in the graph paradigm, which models learnable physical quantities for more realistic real-world object interactions.
π Stay tuned for more updates when the paper is officially submitted and under review!
