Temporal Difference Learning for Model Predictive Control
TD-MPC — a task-oriented latent model used for short-horizon planning, with a learned value function covering the long-term return; the predecessor of TD-MPC2.TD-MPC：用面向任务的潜空间模型做短视距规划，由学习到的价值函数估计长期回报；是 TD-MPC2 的前身。
BibTeX
@misc{hansen2022temporal,
title = {{Temporal Difference Learning for Model Predictive Control}},
author = {Nicklas Hansen and Xiaolong Wang and Hao Su},
year = {2022},
eprint = {2203.04955},
archivePrefix = {arXiv},
primaryClass = {cs.LG},
url = {https://arxiv.org/abs/2203.04955},
}