English

DMC-VB: A Benchmark for Representation Learning for Control with Visual Distractors

Machine Learning 2024-09-30 v1

Abstract

Learning from previously collected data via behavioral cloning or offline reinforcement learning (RL) is a powerful recipe for scaling generalist agents by avoiding the need for expensive online learning. Despite strong generalization in some respects, agents are often remarkably brittle to minor visual variations in control-irrelevant factors such as the background or camera viewpoint. In this paper, we present theDeepMind Control Visual Benchmark (DMC-VB), a dataset collected in the DeepMind Control Suite to evaluate the robustness of offline RL agents for solving continuous control tasks from visual input in the presence of visual distractors. In contrast to prior works, our dataset (a) combines locomotion and navigation tasks of varying difficulties, (b) includes static and dynamic visual variations, (c) considers data generated by policies with different skill levels, (d) systematically returns pairs of state and pixel observation, (e) is an order of magnitude larger, and (f) includes tasks with hidden goals. Accompanying our dataset, we propose three benchmarks to evaluate representation learning methods for pretraining, and carry out experiments on several recently proposed methods. First, we find that pretrained representations do not help policy learning on DMC-VB, and we highlight a large representation gap between policies learned on pixel observations and on states. Second, we demonstrate when expert data is limited, policy learning can benefit from representations pretrained on (a) suboptimal data, and (b) tasks with stochastic hidden goals. Our dataset and benchmark code to train and evaluate agents are available at: https://github.com/google-deepmind/dmc_vision_benchmark.

Keywords

Cite

@article{arxiv.2409.18330,
  title  = {DMC-VB: A Benchmark for Representation Learning for Control with Visual Distractors},
  author = {Joseph Ortiz and Antoine Dedieu and Wolfgang Lehrach and Swaroop Guntupalli and Carter Wendelken and Ahmad Humayun and Guangyao Zhou and Sivaramakrishnan Swaminathan and Miguel Lázaro-Gredilla and Kevin Murphy},
  journal= {arXiv preprint arXiv:2409.18330},
  year   = {2024}
}

Comments

NeurIPS 2024 Datasets and Benchmarks Track. Dataset available at: https://github.com/google-deepmind/dmc_vision_benchmark

R2 v1 2026-06-28T18:58:53.387Z