MOSU2: Long-range Navigation System with Multi-modal Scene Understanding
MOSU2: Long-range Navigation System with Multi-modal Scene Understanding
This research presents MOSU2, a long-range navigation system for robots operating in public outdoor spaces. The system utilizes a two-stage trajectory selection process where an initial stage of contextual reasoning determines when additional geometric and dynamic evidence is required. By integrating vision-language models with metric geometry and motion data, MOSU2 achieves a mean navigation success rate of 94.2% and condition-detection accuracy of 91.4% in real-world tests involving complex scenarios like construction sites and shared roads, representing an improvement over its single-image predecessor.
The paper introduces MOSU2, a two-stage navigation system that combines contextual reasoning with multi-modal scene understanding to improve long-range outdoor navigation success rates from 61.8% to 94.2% in real-world experiments involving pedestrians, cyclists, and vehicles.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
Reported accuracy
72.2%
View original evidence
g and filters candidate trajectories through multi-modal traversability analysis before selecting a trajectory to move the robot. In real-world experiments spanning stop gestures, construction sites, crosswalks, and roads shared with vehicles, cyclists, and pedestrians, MOSU2 raises mean navigation success from 61.8% to 94.2% and condition-detection accuracy from 72.2% to 91.4% over its single-image predecessor, at a
Open source S5Reported accuracy
91.4%
View original evidence
g and filters candidate trajectories through multi-modal traversability analysis before selecting a trajectory to move the robot. In real-world experiments spanning stop gestures, construction sites, crosswalks, and roads shared with vehicles, cyclists, and pedestrians, MOSU2 raises mean navigation success from 61.8% to 94.2% and condition-detection accuracy from 72.2% to 91.4% over its single-image predecessor, at a
Open source S5
- paper: not_reported
- weights: not_reported
- dataset: not_reported
- code: not_reported
What remains unknown
Not established in the collected evidence: Environment, Control, Data origin.
Reported performance applies to the described task. It does not establish general autonomy or deployment readiness.
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
g and filters candidate trajectories through multi-modal traversability analysis before selecting a trajectory to move the robot. In real-world experiments spanning stop gestures, construction sites, crosswalks, and roads shared with vehicles, cyclists, and pedestrians, MOSU2 raises mean navigation success from 61.8% to 94.2% and condition-detection accuracy from 72.2% to 91.4% over its single-image predecessor, at a
Open source S5
es, such as whether an approaching car is slowing or a cyclist will cross the robot's path. Vision-language models (VLMs) can interpret such cues, but they reason over images that lack metric geometry and motion. Supplying this evidence at every step is too slow for real-time navigation, and hand-crafted triggers cannot anticipate the open-ended variety of outdoor situations. We present MOSU2, a long-range navigation
Open source S3
bout 2.2 times its mean decision latency. Project website: https://jinggm.github.io/mosu2/
Open source S6
Source：arXiv Robotics — research abstracts · arxiv.org