Challenges of real-world reinforcement learning: definitions, benchmarks and analysis

780 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

A review on Deep Reinforcement Learning for Fluid Mechanics (2019, neutral): “[36, 37, 38] for comprehensive domain-agnostic surveys.”
Deep Reinforcement Learning for Robotics: A Survey of Real-World Successes (2024, background): “Most existing surveys on RL for robotics do not explicitly address this topic, e.g., Dulac-Arnold et al. (9) discuss the general challenges of real-world RL not specific to robotics, and Ibarz et al. (10) list open challenges of DRL unique to real-world robotics settings but based on case studies dr”
Challenges in Deploying Machine Learning: A Survey of Case Studies (2020, background): “This point is emphasized further by the experiences from the ield of reinforcement learning [40], where use of simulations is a de-facto standard for training agents.”
AI Autonomy: Self-initiated Open-world Continual Learning and Adaptation (2022, background): “As indicated earlier, this is especially challenging in a physical environment (Dulac-Arnold et al. 2021).”
Is Conditional Generative Modeling all you need for Decision-Making? (2022, background): “This is particularly useful in real-world settings where interacting with the environment is not always possible, and exploratory decisions can have fatal consequences (Dulac-Arnold et al., 2021).”
Self-Initiated Open World Learning for Autonomous AI Agents (2021, background): “This task is especially challenging in a real-world physical environment (Dulac-Arnold et al. 2021).”