Walk the Random Walk: Learning to Discover and Reach Goals Without Supervision - Inria - Institut national de recherche en sciences et technologies du numérique Accéder directement au contenu
Communication Dans Un Congrès Année : 2022

Walk the Random Walk: Learning to Discover and Reach Goals Without Supervision

Résumé

Learning a diverse set of skills by interacting with an environment without any external supervision is an important challenge. In particular, obtaining a goal-conditioned agent that can reach any given state is useful in many applications. We propose a novel method for training such a goal-conditioned agent without any external rewards or any domain knowledge. We use random walk to train a reachability network that predicts the similarity between two states. This reachability network is then used in building goal memory containing past observations that are diverse and well-balanced. Finally, we train a goal-conditioned policy network with goals sampled from the goal memory and reward it by the reachability network and the goal memory. All the components are kept updated throughout training as the agent discovers and learns new goals. We apply our method to a continuous control navigation and robotic manipulation tasks.
Fichier principal
Vignette du fichier
26_walk_the_random_walk_learning_.pdf (1.33 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04107096 , version 1 (25-05-2023)

Licence

Paternité

Identifiants

Citer

Lina Mezghani, Sainbayar Sukhbaatar, Piotr Bojanowski, Karteek Alahari. Walk the Random Walk: Learning to Discover and Reach Goals Without Supervision. ICLR 2022 - Workshop on Agent Learning in Open-Endedness (ALOE), Apr 2022, Online (virtual), Unknown Region. ⟨hal-04107096⟩
37 Consultations
14 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More