Megaverse: Simulating Embodied Agents at One Million Experiences per Second

Aleksei Petrenko, Erik Wijmans, Brennan Shacklett, Vladlen Koltun
Proceedings of the 38th International Conference on Machine Learning, PMLR 139:8556-8566, 2021.

Abstract

We present Megaverse, a new 3D simulation platform for reinforcement learning and embodied AI research. The efficient design of our engine enables physics-based simulation with high-dimensional egocentric observations at more than 1,000,000 actions per second on a single 8-GPU node. Megaverse is up to 70x faster than DeepMind Lab in fully-shaded 3D scenes with interactive objects. We achieve this high simulation performance by leveraging batched simulation, thereby taking full advantage of the massive parallelism of modern GPUs. We use Megaverse to build a new benchmark that consists of several single-agent and multi-agent tasks covering a variety of cognitive challenges. We evaluate model-free RL on this benchmark to provide baselines and facilitate future research.

Cite this Paper

BibTeX
@InProceedings{pmlr-v139-petrenko21a, title = {Megaverse: Simulating Embodied Agents at One Million Experiences per Second}, author = {Petrenko, Aleksei and Wijmans, Erik and Shacklett, Brennan and Koltun, Vladlen}, booktitle = {Proceedings of the 38th International Conference on Machine Learning}, pages = {8556--8566}, year = {2021}, editor = {Meila, Marina and Zhang, Tong}, volume = {139}, series = {Proceedings of Machine Learning Research}, month = {18--24 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v139/petrenko21a/petrenko21a.pdf}, url = {https://proceedings.mlr.press/v139/petrenko21a.html}, abstract = {We present Megaverse, a new 3D simulation platform for reinforcement learning and embodied AI research. The efficient design of our engine enables physics-based simulation with high-dimensional egocentric observations at more than 1,000,000 actions per second on a single 8-GPU node. Megaverse is up to 70x faster than DeepMind Lab in fully-shaded 3D scenes with interactive objects. We achieve this high simulation performance by leveraging batched simulation, thereby taking full advantage of the massive parallelism of modern GPUs. We use Megaverse to build a new benchmark that consists of several single-agent and multi-agent tasks covering a variety of cognitive challenges. We evaluate model-free RL on this benchmark to provide baselines and facilitate future research.} }
Endnote
%0 Conference Paper %T Megaverse: Simulating Embodied Agents at One Million Experiences per Second %A Aleksei Petrenko %A Erik Wijmans %A Brennan Shacklett %A Vladlen Koltun %B Proceedings of the 38th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2021 %E Marina Meila %E Tong Zhang %F pmlr-v139-petrenko21a %I PMLR %P 8556--8566 %U https://proceedings.mlr.press/v139/petrenko21a.html %V 139 %X We present Megaverse, a new 3D simulation platform for reinforcement learning and embodied AI research. The efficient design of our engine enables physics-based simulation with high-dimensional egocentric observations at more than 1,000,000 actions per second on a single 8-GPU node. Megaverse is up to 70x faster than DeepMind Lab in fully-shaded 3D scenes with interactive objects. We achieve this high simulation performance by leveraging batched simulation, thereby taking full advantage of the massive parallelism of modern GPUs. We use Megaverse to build a new benchmark that consists of several single-agent and multi-agent tasks covering a variety of cognitive challenges. We evaluate model-free RL on this benchmark to provide baselines and facilitate future research.
APA
Petrenko, A., Wijmans, E., Shacklett, B. & Koltun, V.. (2021). Megaverse: Simulating Embodied Agents at One Million Experiences per Second. Proceedings of the 38th International Conference on Machine Learning, in Proceedings of Machine Learning Research 139:8556-8566 Available from https://proceedings.mlr.press/v139/petrenko21a.html.

Related Material