FetchBench: A Simulation Benchmark for Robot Fetching

Beining Han, Meenal Parakh, Derek Geng, Jack A Defay, Gan Luyang, Jia Deng
Proceedings of The 8th Conference on Robot Learning, PMLR 270:3053-3071, 2025.

Abstract

Fetching, which includes approaching, grasping, and retrieving, is a critical challenge for robot manipulation tasks. Existing methods primarily focus on table-top scenarios, which do not adequately capture the complexities of environments where both grasping and planning are essential. To address this gap, we propose a new benchmark FetchBench, featuring diverse procedural scenes that integrate both grasping and motion planning challenges. Additionally, FetchBench includes a data generation pipeline that collects successful fetch trajectories for use in imitation learning methods. We implement multiple baselines from the traditional sense-plan-act pipeline to end-to-end behavior models. Our empirical analysis reveals that these methods achieve a maximum success rate of only 20%, indicating substantial room for improvement. Additionally, we identify key bottlenecks within the sense-plan-act pipeline and make recommendations based on the systematic analysis.

Cite this Paper

BibTeX
@InProceedings{pmlr-v270-han25a, title = {FetchBench: A Simulation Benchmark for Robot Fetching}, author = {Han, Beining and Parakh, Meenal and Geng, Derek and Defay, Jack A and Luyang, Gan and Deng, Jia}, booktitle = {Proceedings of The 8th Conference on Robot Learning}, pages = {3053--3071}, year = {2025}, editor = {Agrawal, Pulkit and Kroemer, Oliver and Burgard, Wolfram}, volume = {270}, series = {Proceedings of Machine Learning Research}, month = {06--09 Nov}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v270/main/assets/han25a/han25a.pdf}, url = {https://proceedings.mlr.press/v270/han25a.html}, abstract = {Fetching, which includes approaching, grasping, and retrieving, is a critical challenge for robot manipulation tasks. Existing methods primarily focus on table-top scenarios, which do not adequately capture the complexities of environments where both grasping and planning are essential. To address this gap, we propose a new benchmark FetchBench, featuring diverse procedural scenes that integrate both grasping and motion planning challenges. Additionally, FetchBench includes a data generation pipeline that collects successful fetch trajectories for use in imitation learning methods. We implement multiple baselines from the traditional sense-plan-act pipeline to end-to-end behavior models. Our empirical analysis reveals that these methods achieve a maximum success rate of only 20%, indicating substantial room for improvement. Additionally, we identify key bottlenecks within the sense-plan-act pipeline and make recommendations based on the systematic analysis.} }
Endnote
%0 Conference Paper %T FetchBench: A Simulation Benchmark for Robot Fetching %A Beining Han %A Meenal Parakh %A Derek Geng %A Jack A Defay %A Gan Luyang %A Jia Deng %B Proceedings of The 8th Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2025 %E Pulkit Agrawal %E Oliver Kroemer %E Wolfram Burgard %F pmlr-v270-han25a %I PMLR %P 3053--3071 %U https://proceedings.mlr.press/v270/han25a.html %V 270 %X Fetching, which includes approaching, grasping, and retrieving, is a critical challenge for robot manipulation tasks. Existing methods primarily focus on table-top scenarios, which do not adequately capture the complexities of environments where both grasping and planning are essential. To address this gap, we propose a new benchmark FetchBench, featuring diverse procedural scenes that integrate both grasping and motion planning challenges. Additionally, FetchBench includes a data generation pipeline that collects successful fetch trajectories for use in imitation learning methods. We implement multiple baselines from the traditional sense-plan-act pipeline to end-to-end behavior models. Our empirical analysis reveals that these methods achieve a maximum success rate of only 20%, indicating substantial room for improvement. Additionally, we identify key bottlenecks within the sense-plan-act pipeline and make recommendations based on the systematic analysis.
APA
Han, B., Parakh, M., Geng, D., Defay, J.A., Luyang, G. & Deng, J.. (2025). FetchBench: A Simulation Benchmark for Robot Fetching. Proceedings of The 8th Conference on Robot Learning, in Proceedings of Machine Learning Research 270:3053-3071 Available from https://proceedings.mlr.press/v270/han25a.html.

Related Material