Adversarial Actor-Critic Method for Task and Motion Planning Problems Using Planning Experience

Abstract

We propose an actor-critic algorithm that uses past planning experience to improve the efficiency of solving robot task-and-motion planning (TAMP) problems. TAMP planners search for goal-achieving sequences of high-level operator instances specified by both discrete and continuous parameters. Our algorithm learns a policy for selecting the continuous parameters during search, using a small training set generated from the search trees of previously solved instances. We also introduce a novel fixed-length vector representation for world states with varying numbers of objects with different shapes, based on a set of key robot configurations. We demonstrate experimentally that our method learns more efficiently from less data than standard reinforcementlearning approaches and that using a learned policy to guide a planner results in the improvement of planning efficiency.

Cite

Text

Kim et al. "Adversarial Actor-Critic Method for Task and Motion Planning Problems Using Planning Experience." AAAI Conference on Artificial Intelligence, 2019. doi:10.1609/AAAI.V33I01.33018017

Markdown

[Kim et al. "Adversarial Actor-Critic Method for Task and Motion Planning Problems Using Planning Experience." AAAI Conference on Artificial Intelligence, 2019.](https://mlanthology.org/aaai/2019/kim2019aaai-adversarial/) doi:10.1609/AAAI.V33I01.33018017

BibTeX

@inproceedings{kim2019aaai-adversarial,
  title     = {{Adversarial Actor-Critic Method for Task and Motion Planning Problems Using Planning Experience}},
  author    = {Kim, Beomjoon and Kaelbling, Leslie Pack and Lozano-Pérez, Tomás},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year      = {2019},
  pages     = {8017-8024},
  doi       = {10.1609/AAAI.V33I01.33018017},
  url       = {https://mlanthology.org/aaai/2019/kim2019aaai-adversarial/}
}