Towards Automata-Based Abstraction of Goals in Hierarchical Reinforcement Learning - Archive ouverte HAL Access content directly
Conference Papers Year :

Towards Automata-Based Abstraction of Goals in Hierarchical Reinforcement Learning

Abstract

Hierarchical Reinforcement Learning (HRL) offers potential benefits for solving long horizon tasks, generally unhandled by standard Reinforcement Learning (RL) techniques, by decomposing the problem and combining simple policies to achieve the goal. They are however still held back by the curse of dimensionality and the ambiguity of selected subtasks. We explore relevant approaches in HRL while highlighting the key challenges of goal representation, high-level planning and propose a research outline tackling them.
Fichier principal
Vignette du fichier
bare_conf.pdf (57.49 Ko) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-03600799 , version 1 (07-03-2022)

Identifiers

  • HAL Id : hal-03600799 , version 1

Cite

Mehdi Zadem, Sergio Mover, Sao Mai Nguyen, Sylvie Putot. Towards Automata-Based Abstraction of Goals in Hierarchical Reinforcement Learning. Intrinsically Motivated Open-ended Learning IMOL 2022, Apr 2022, Tübingen, Germany. ⟨hal-03600799⟩
128 View
132 Download

Share

Gmail Facebook Twitter LinkedIn More