Goal Space Abstraction in Hierarchical Reinforcement Learning via Set-Based Reachability Analysis

Mehdi Zadem (1, 2) , Sergio Mover (1) , Sao Mai Nguyen (3, 4, 2, 5)

1 LIX - Laboratoire d'informatique de l'École polytechnique [Palaiseau]
2 U2IS - Unité d'Informatique et d'Ingénierie des Systèmes
3 IMT Atlantique - INFO - Département Informatique
4 Lab-STICC_RAMBO - Equipe Robot interaction, Ambient system, Machine learning, Behaviour, Optimization
5 Flowers - Flowing Epigenetic Robots and Systems

Open-ended learning benefits immensely from the use of symbolic methods for goal representation as they offer ways to structure knowledge for efficient and transferable learning. However, the existing Hierarchical Reinforcement Learning (HRL) approaches relying on symbolic reasoning are often limited as they require a manual goal representation. The challenge in autonomously discovering a symbolic goal representation is that it must preserve critical information, such as the environment dynamics. In this paper, we propose a developmental mechanism for goal discovery via an emergent representation that abstracts (i.e., groups together) sets of environment states that have similar roles in the task. We introduce a Feudal HRL algorithm that concurrently learns both the goal representation and a hierarchical policy. The algorithm uses symbolic reachability analysis for neural networks to approximate the transition relation among sets of states and to refine the goal representation. We evaluate our approach on complex navigation tasks, showing the learned representation is interpretable, transferrable and results in data efficient learning.

Keywords: Reinforcement Leaning RL; Hierarchical reinforcement learning; Open-ended learning; Intrinsic motivation; Curriculum learning; deep learning; Deep reinforcement learning DRL

Domains: Machine Learning [cs.LG]; Artificial Intelligence [cs.AI]

Mehdi Zadem, Sergio Mover, Sao Mai Nguyen. Goal Space Abstraction in Hierarchical Reinforcement Learning via Set-Based Reachability Analysis. ICDL 2023 - 22nd IEEE International Conference on Development and Learning, Nov 2023, Macau, China. pp.423-428, ⟨10.1109/ICDL55364.2023.10364473⟩. ⟨hal-04401878⟩