研究人员针对传统图神经网络(GNNs)在复杂场景和长期预测中的局限性，开发了高阶拓扑物理信息网络(HOPNet)，通过组合复形(Combinatorial Complex)表示和物理约束消息传递机制，显著提升了多刚体碰撞动力学模拟的精度和泛化能力。该框架在MOVi系列数据集测试中 ...

A framework based on advanced AI techniques can solve complex, computationally intensive problems faster and in a more more scalable way than state-of-the-art methods, according to a new study. A ...