MineRL: A Large-Scale Dataset of Minecraft Demonstrations

By William H. Guss, Brandon Houghton, Nicholay Topin, Phillip Wang, Cayden R. Codel, Manuela M. Veloso, Ruslan Salakhutdinov. International Joint Conference on Artificial Intelligence, 2019

305 citations recorded by the source corpus.

The sample inefficiency of standard deep reinforcement learning methods precludes their application to many real-world problems. Methods which leverage human demonstrations require fewer samples but have been researched less. As demonstrated in the computer vision and natural language processing communities, large-scale datasets have the capacity to facilitate research by serving as an experimental and benchmarking platform for new methods. However, existing datasets compatible with reinforcement learning simulators do not have sufficient scale, structure, and quality to enable the further development and evaluation of methods focused on using human examples. Therefore, we introduce a comprehensive, large-scale, simulator-paired dataset of human demonstrations: MineRL. The dataset consists of over 60 million automatically annotated state-action pairs across a variety of related tasks in Minecraft, a dynamic, 3D, open-world environment. We present a novel data collection scheme which allows for the ongoing introduction of new tasks and the gathering of complete state information suitable for a variety of methods. We demonstrate the hierarchality, diversity, and scale of the MineRL dataset. Further, we show the difficulty of the Minecraft domain along with the potential of MineRL in developing techniques to solve key research challenges within it.

Subfields: Artificial Intelligence, Control and Systems Engineering

View Paper: https://www.semanticscholar.org/paper/2c87b6b48dba6315a6581098861f8bf2994799f6

How researchers see this paper

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

In Artificial Intelligence

In Control and Systems Engineering

Citations in the corpus

Video PreTraining (VPT): Learning to Act by Watching Unlabeled Online Videos (2022, background): “[27] William H Guss, Brandon Houghton, Nicholay Topin, Phillip Wang, Cayden Codel, Manuela Veloso, and Ruslan Salakhutdinov.”
AndroidEnv: A Reinforcement Learning Platform for Android (2021, methodology): “Other methods leverage prior human demonstration or prior knowledge (Abel et al., 2015, 2016; Frazier and Riedl, 2019; Guss et al., 2019; Shu et al., 2017).”
Describe, Explain, Plan and Select: Interactive Planning with Large Language Models Enables Open-World Multi-Task Agents (2023, methodology): “Considering how long it typically takes human players to complete each task as a ballpark [14], we set different maximum episode steps for different meta tasks from 3000 (for easiest Basic tasks) to 12000 (for the hardest Challenge tasks).”
Atari-HEAD: Atari Human Eye-Tracking and Demonstration Dataset (2019, background): “Recently, researchers have spent signiﬁcant effort in building large-scale datasets of human demonstrations in various tasks, including driving (Yu et al. 2018), playing Minecraft (Guss et al. 2019), and manipulating simulated robots (Mandlekar et al. 2018).”
Flexible Diffusion Modeling of Long Videos (2022, methodology): “We present results on three datasets: GQN-Mazes [10], in which videos are 300 frames long; MineRL Navigate [13, 26] (which we will from now on refer to as simply MineRL), in which videos are 500 frames long; and the CARLA Town01 dataset we release, for which videos are 1000 frames long.”
MineDojo: Building Open-Ended Embodied Agents with Internet-Scale Knowledge (2022, methodology): “Based on Malmo, MineRL [48] provides a codebase and human play trajectories for the annual Diamond Challenge at NeurIPS [47, 49, 62].”
Language Agent Tree Search Unifies Reasoning Acting and Planning in Language Models (2023, methodology): “Using planning-based prompting methods like LATS in environments like Minecraft (Fan et al., 2022) and more reasoning benchmarks would be interesting avenues for future work.”
The MineRL 2020 Competition on Sample Efficient Reinforcement Learning using Human Priors (2021, methodology): “To that end, we center our proposed competition around techniques which leverage the MineRL dataset [13].”
Jelly Bean World: A Testbed for Never-Ending Learning (2020, background): “, 2016), (Guss et al., 2019) ∼ 7 3 3 7 7 Tasks have a pre-specified time limit but that is typically very long Stationary rewards and map generation is based on Perlin noise Supports 6 complex tasks but also allows for new ones Agents observe the game video frames and the game state Modifying the ta”
Do Embodied Agents Dream of Pixelated Sheep?: Embodied Decision Making using Language Guided World Modelling (2023, background): “We found Minedojo preferable to MineRL (Guss et al., 2019), due to a reduced tendency to crash when running many parallel environment instances.”
Minigrid & Miniworld: Modular & Customizable Reinforcement Learning Environments for Goal-Oriented Tasks (2023, methodology): “Another 3D simulation library with a similar purpose is DeepMind Lab [2], but given that the game engine is written in C and the levels are written using Lua, there is a steep learning curve for customizing it.”