# Variational Dynamic for Self-Supervised Exploration in Deep Reinforcement Learning

arXiv: https://arxiv.org/abs/2010.08755 · HTML: https://mlsanity.com/paper/2010.08755/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 17 Oct 2020.

## Built on by (most cited first, 9 of 9)

- [Exploration in Deep Reinforcement Learning: From Single-Agent to Multiagent Domain](https://mlsanity.com/paper/2109.06668/) (2021)
- [Principled Exploration via Optimistic Bootstrapping and Backward Induction](https://mlsanity.com/paper/2105.06022/) (2021)
- [Behavior Contrastive Learning for Unsupervised Skill Discovery](https://mlsanity.com/paper/2305.04477/) (2023)
- [Curiosity-Driven Exploration via Latent Bayesian Surprise](https://mlsanity.com/paper/2104.07495/) (2021)
- [Dynamic Bottleneck for Robust Self-Supervised Exploration](https://mlsanity.com/paper/2110.10735/) (2021)
- [ViTaS: Visual Tactile Soft Fusion Contrastive Learning for Visuomotor Learning](https://mlsanity.com/paper/2602.11643/) (2026)
- [Temporal Representations for Exploration: Learning Complex Exploratory Behavior without Extrinsic Rewards](https://mlsanity.com/paper/2603.02008/) (2026)
- [Goal Exploration via Adaptive Skill Distribution for Goal-Conditioned Reinforcement Learning](https://mlsanity.com/paper/2404.12999/) (2024)
- [Feasibility-Aware Pessimistic Estimation: Toward Long-Horizon Safety in Offline RL](https://mlsanity.com/paper/2505.08179/) (2025)

## Abstract (arXiv)

Efficient exploration remains a challenging problem in reinforcement learning, especially for tasks where extrinsic rewards from environments are sparse or even totally disregarded. Significant advances based on intrinsic motivation show promising results in simple environments but often get stuck in environments with multimodal and stochastic dynamics. In this work, we propose a variational dynamic model based on the conditional variational inference to model the multimodality and stochasticity. We consider the environmental state-action transition as a conditional generative process by generating the next-state prediction under the condition of the current state, action, and latent variable, which provides a better understanding of the dynamics and leads a better performance in exploration. We derive an upper bound of the negative log-likelihood of the environmental transition and use such an upper bound as the intrinsic reward for exploration, which allows the agent to learn skills by self-supervised exploration without observing extrinsic rewards. We evaluate the proposed method on several image-based simulation tasks and a real robotic manipulating task. Our method outperforms several state-of-the-art environment model-based exploration approaches.