arXiv Robotics — research abstracts· Hongyu Ding, Kanyu Xu, Mingjia Chen, Zixuan Chen, Jieqi Shi, Jing Huo, Yifan Zhang, Yang Gao·· 11 hours agoEditorial score65
HM3D-M3: A Benchmark for Multi-Agent, Multi-Modal, Multi-Object Goal Navigation
HM3D-M3: A Benchmark for Multi-Agent, Multi-Modal, Multi-Object Goal Navigation
Summary
This research introduces HM3D-M3, a benchmark for multi-agent, multi-modal, and multi-object goal navigation. The benchmark features 1,000 validation episodes across 3,926 goals from 379 categories, along with training samples and team-level metrics. Two baselines are proposed, achieving 38-41% success rates, with plans to deploy one on a heterogeneous robot team.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
Source：arXiv Robotics — research abstracts · arxiv.org