English

Goal-Directed Planning by Reinforcement Learning and Active Inference

Machine Learning 2021-06-23 v2 Artificial Intelligence Robotics

Abstract

What is the difference between goal-directed and habitual behavior? We propose a novel computational framework of decision making with Bayesian inference, in which everything is integrated as an entire neural network model. The model learns to predict environmental state transitions by self-exploration and generating motor actions by sampling stochastic internal states z{z}. Habitual behavior, which is obtained from the prior distribution of z{z}, is acquired by reinforcement learning. Goal-directed behavior is determined from the posterior distribution of z{z} by planning, using active inference which optimizes the past, current and future z{z} by minimizing the variational free energy for the desired future observation constrained by the observed sensory sequence. We demonstrate the effectiveness of the proposed framework by experiments in a sensorimotor navigation task with camera observations and continuous motor actions.

Keywords

Cite

@article{arxiv.2106.09938,
  title  = {Goal-Directed Planning by Reinforcement Learning and Active Inference},
  author = {Dongqi Han and Kenji Doya and Jun Tani},
  journal= {arXiv preprint arXiv:2106.09938},
  year   = {2021}
}

Comments

Work in progress

R2 v1 2026-06-24T03:20:50.335Z