CusADi2: Latency-Optimized Batch Computation of Robot Kinematics, Jacobians, and Hessians on the GPU
Evaluates batches of kinematics, Jacobians and Hessians on the GPU with low wall-clock latency, targeting online sampling- and gradient-based trajectory optimization.在 GPU 上以低延迟批量计算运动学、雅可比与海森矩阵，面向在线运行的采样式与梯度式轨迹优化。
BibTeX
@misc{viljoen2026cusadi,
title = {{CusADi2: Latency-Optimized Batch Computation of Robot Kinematics, Jacobians, and Hessians on the GPU}},
author = {John Viljoen and Lasse Peters and Masayoshi Tomizuka and Negar Mehr},
year = {2026},
eprint = {2610.25476},
archivePrefix = {arXiv},
primaryClass = {cs.RO},
url = {https://arxiv.org/abs/2610.25476},
}