提出显式凸可行区,高效处理变形体碰撞。
Convex Collision-Free Regions

- 为每个顶点构建独立凸可行区,预定义非穿透位移范围。
- 支持主/次碰撞及零维接触,无需依赖非线性优化。
- 兼容XPBD等框架,适合大规模并行模拟场景。
凸无碰撞区域(CCFR)是一种显式表示局部凸可行区域的碰撞处理方法,通过周围网格基元配置(包括边-边、顶点-面交互)构建,表征当前构型下可接受的非穿透顶点位移。现有变形体模拟中的碰撞处理多依赖隐式可行性表示,导致二次碰撞和零维接触鲁棒性差,或与特定非线性优化方案强耦合。本方法为每个顶点独立构建可行区域,预先定义,不仅涵盖主碰撞,还可自然处理次级碰撞与零维接触,实现高度可扩展且可并行的碰撞处理。这些可行区域独立于物理接触响应模型编码几何非穿透约束。CCFR不依赖非线性优化,兼容如扩展位置基动力学(XPBD)等不显式维护内部可行性的仿真框架。在布料、发丝、线状物、粒子系统及零维接触场景中验证了其通用性与高效性。
原文摘要 · Abstract (English)
Convex Collision-Free Regions (CCFR) is a collision handling method that explicitly represents local convex feasible regions to enforce non-penetration. Each feasible region is constructed from surrounding mesh primitive configurations, including edge-edge and vertex-face interactions. The resulting convex region represents admissible non-penetrating vertex displacements at the current configuration. Existing collision handling methods for deformable body simulation have largely relied on implicit representations of feasibility, resulting in either compromised robustness for secondary collisions and codimensional contacts or tight coupling with specific nonlinear optimization schemes. Our formulation constructs feasible regions independently for each vertex, defined prior to penetration, inherently accounts not only for primary collisions but also for secondary collisions and codimensional contacts, enabling highly scalable and parallelizable collision handling. These feasible regions encode geometric non-penetration constraints independently of physical contact response models. CCFR does not rely on nonlinear optimization and is compatible with simulation frameworks such as Extended Position-Based Dynamics (XPBD) that do not explicitly maintain interior feasibility during iterative updates. The effectiveness of CCFR is demonstrated across cloth, hair, wire, particle systems, and codimensional contact scenarios, showing versatile and efficient collision handling.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品;中文卡片由大模型生成,请以原文为准。