PrismPlan: A Shared GPU Substrate for Motion Planning Across Paradigms
Builds batched GPU collision checking and forward kinematics once and runs sampling, search and optimization planners on top, so the three paradigms can be compared fairly.一次性实现批量 GPU 碰撞检测与正运动学，在其上运行采样、搜索和优化三类规划器，使三种范式可以公平比较。
BibTeX
@misc{wu2026prismplan,
title = {{PrismPlan: A Shared GPU Substrate for Motion Planning Across Paradigms}},
author = {Hong Wu and ChaoXuan Liu and ShuJie Wang},
year = {2026},
eprint = {2610.23884},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.23884},
}