# Self-Supervised Policy Adaptation during Deployment

arXiv: https://arxiv.org/abs/2007.04309 · HTML: https://mlsanity.com/paper/2007.04309/

## Key facts

- Cited by 123 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 8 Jul 2020.

## Built on by (most cited first, 30 of 123)

- [RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control](https://mlsanity.com/paper/2307.15818/) (2023)
- [Temporal Difference Learning for Model Predictive Control](https://mlsanity.com/paper/2203.04955/) (2022)
- [A Comprehensive Survey on Test-Time Adaptation under Distribution Shifts](https://mlsanity.com/paper/2303.15361/) (2023)
- [Learning to (Learn at Test Time): RNNs with Expressive Hidden States](https://mlsanity.com/paper/2407.04620/) (2024)
- [Tool Learning with Foundation Models](https://mlsanity.com/paper/2304.08354/) (2023)
- [Test-Time Training with Masked Autoencoders](https://mlsanity.com/paper/2209.07522/) (2022)
- [A Survey on Self-supervised Learning: Algorithms, Applications, and Future Trends](https://mlsanity.com/paper/2301.05712/) (2023)
- [Scaling Robot Learning with Semantically Imagined Experience](https://mlsanity.com/paper/2302.11550/) (2023)
- [A Survey of Zero-shot Generalisation in Deep Reinforcement Learning](https://mlsanity.com/paper/2111.09794/) (2021)
- [Data Scaling Laws in Imitation Learning for Robotic Manipulation](https://mlsanity.com/paper/2410.18647/) (2024)
- [Stabilizing Deep Q-Learning with ConvNets and Vision Transformers under Data Augmentation](https://mlsanity.com/paper/2107.00644/) (2021)
- [The Distracting Control Suite -- A Challenging Benchmark for Reinforcement Learning from Pixels](https://mlsanity.com/paper/2101.02722/) (2021)
- [Test-Time Adaptation via Conjugate Pseudo-labels](https://mlsanity.com/paper/2207.09640/) (2022)
- [End-to-End Test-Time Training for Long Context](https://mlsanity.com/paper/2512.23675/) (2025)
- [Pre-Trained Image Encoder for Generalizable Visual Reinforcement Learning](https://mlsanity.com/paper/2212.08860/) (2022)
- [RobustNav: Towards Benchmarking Robustness in Embodied Navigation](https://mlsanity.com/paper/2106.04531/) (2021)
- [Learning Generalizable Dexterous Manipulation from Human Grasp Affordance](https://mlsanity.com/paper/2204.02320/) (2022)
- [Look Closer: Bridging Egocentric and Third-Person Views with Transformers for Robotic Manipulation](https://mlsanity.com/paper/2201.07779/) (2022)
- [Visual Reinforcement Learning with Self-Supervised 3D Representations](https://mlsanity.com/paper/2210.07241/) (2022)
- [SECANT: Self-Expert Cloning for Zero-Shot Generalization of Visual Policies](https://mlsanity.com/paper/2106.09678/) (2021)
- [Learning Navigational Visual Representations with Semantic Map Supervision](https://mlsanity.com/paper/2307.12335/) (2023)
- [Augmented World Models Facilitate Zero-Shot Dynamics Generalization From a Single Offline Environment](https://mlsanity.com/paper/2104.05632/) (2021)
- [Centaur: Robust End-to-End Autonomous Driving with Test-Time Training](https://mlsanity.com/paper/2503.11650/) (2025)
- [Finetuning Offline World Models in the Real World](https://mlsanity.com/paper/2310.16029/) (2023)
- [Learning to Manipulate Anywhere: A Visual Generalizable Framework For Reinforcement Learning](https://mlsanity.com/paper/2407.15815/) (2024)
- [RoMA: Robust Model Adaptation for Offline Model-based Optimization](https://mlsanity.com/paper/2110.14188/) (2021)
- [What Makes Pre-Trained Visual Representations Successful for Robust Manipulation?](https://mlsanity.com/paper/2312.12444/) (2023)
- [Learning to Drive by Watching YouTube Videos: Action-Conditioned Contrastive Policy Pretraining](https://mlsanity.com/paper/2204.02393/) (2022)
- [Test-Time Training on Video Streams](https://mlsanity.com/paper/2307.05014/) (2023)
- [Test-Time Personalization with a Transformer for Human Pose Estimation](https://mlsanity.com/paper/2107.02133/) (2021)

## Abstract (arXiv)

In most real world scenarios, a policy trained by reinforcement learning in one environment needs to be deployed in another, potentially quite different environment. However, generalization across different environments is known to be hard. A natural solution would be to keep training after deployment in the new environment, but this cannot be done if the new environment offers no reward signal. Our work explores the use of self-supervision to allow the policy to continue training after deployment without using any rewards. While previous methods explicitly anticipate changes in the new environment, we assume no prior knowledge of those changes yet still obtain significant improvements. Empirical evaluations are performed on diverse simulation environments from DeepMind Control suite and ViZDoom, as well as real robotic manipulation tasks in continuously changing environments, taking observations from an uncalibrated camera. Our method improves generalization in 31 out of 36 environments across various tasks and outperforms domain randomization on a majority of environments.