Event
RMMBench Robot Mobile Manipulation Benchmark Released
1 reports · 1 sources
Huapeng Li and other authors propose and release RMMBench, a mobile manipulation evaluation benchmark that requires robots to understand language instructions and execute long-horizon tasks in continuous space. The benchmark integrates high-level and low-level embodied tasks into a unified framework, building a "navigation-manipulation" task suite covering 70 typical scenarios, extending from local manipulation to long-horizon composite navigation. The authors state that evaluation results show leading vision-language models (VLMs) still face notable challenges in spatial grounding for mobile manipulation tasks, and their spatial perception capabilities in long-horizon interactions need strengthening. This work has been accepted by CoRL 2026.
Report timeline
- RMMBench: A Comprehensive Benchmark for Robotic Mobile Manipulation
The authors propose RMMBench, a mobile manipulation benchmark that requires robots to understand language instructions and execute long-horizon tasks in continuous space, with a "navigation-manipulation" task suite covering 70 typical task scenarios. Results show that leading vision-language models (VLMs) still face notable challenges in spatial localization for mobile manipulation tasks, and their spatial perception in long-horizon interactions needs strengthening. The work has been accepted to CoRL 2026.
arXiv cs.RO ·