Grounding Object-to-Object Hazard Reasoning into 3D Risk Fields for Task-Conditioned Safety
Organizations: Department of Computer Science, McGill University, Montreal, Canada. · Department of Electrical and Computer Engineering, McGill University, Montreal, Canada.
Abstract
Safety in robotics does not solely depend on geometric constraints, but also on semantics: the given task, environment properties, and interactions between objects in the scene. We present a framework based on vision-language models (VLMs) to infer task- and scene-dependent safety constraints for robot manipulation. We provide the VLM a structured safety tool based on an RGB-D observation and a natural-language task instruction that first decides whether the task should be rejected since its execution may result in a non-recoverable hazardous outcome. Otherwise, it identifies preventable object-pair hazards and specifies the spatial relations that must be avoided. Open-vocabulary segmentation and depth information are then used to ground the detected objects in 3D and construct continuous risk fields that represent the corresponding hazardous regions. These fields are integrated with an obstacle avoidance module that corrects nominal robot motions during execution while accounting for the risk field. We evaluate the framework on preventable hazards, tasks requiring rejection, and safe variants in simulation and the real-world. Project website: https://vlm-hazard-riskfield.github.io
Figures & tables
| VLM | Cost | Prompt | Tool (ours) | ||||
|---|---|---|---|---|---|---|---|
| Rel. | F1 | Miss | F.rej. | F1 | Miss | F.rej. | |
| GPT-4o | 1.0 | 82.8 | 25.0 | 79.8 | 86.2 | 7.5 | 0.0 |
| Claude Haiku 4.5 | 0.7 | 23.2 | 73.9 | 25.6 | 38.1 | 60.3 | 9.3 |
| Gemini ER 2 | 1.1 | 91.1 | 13.2 | 0.0 | 73.8 | 41.5 | 0.0 |
| Claude Sonnet 5 | 1.6 | 60.6 | 27.0 | 0.0 | 77.3 | 23.9 | 0.0 |
| Gemini 3.1 Pro | 1.7 | 99.3 | 0.7 | 10.9 | 86.6 | 23.6 | 0.8 |
| [6pt][0pt] | Preventable hazard | Safe variant | Overall | |||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| [6pt][0pt] | Drawer-plant | Laptop-Cup | Steak-Dripping | Drawer-Safe | Laptop-Cube | Steak-Safe | Average | |||||||||||
| [6pt][6pt] Method | SR | SSR | Viol. | SR | SSR | Viol. | SR | SSR | Viol. | SR | SSR | SR | SSR | SR | SSR | SR | SSR | Viol. |
| No context | 33.3 [-1.5pt] 15.3 | 0.0 [-1.5pt] 0.0 | 4.1 [-1.5pt] 1.3 | 63.3 [-1.5pt] 5.8 | 46.7 [-1.5pt] 5.8 | 3.8 [-1.5pt] 2.8 | 100.0 [-1.5pt] 0.0 | 0.0 [-1.5pt] 0.0 | 11.4 [-1.5pt] 0.1 | 26.7 [-1.5pt] 5.8 | 26.7 [-1.5pt] 5.8 | 100.0 [-1.5pt] 0.0 | 100.0 [-1.5pt] 0.0 | 100.0 [-1.5pt] 0.0 | 100.0 [-1.5pt] 0.0 | 70.6 [-1.5pt] 4.2 | 45.6 [-1.5pt] 1.0 | 6.5 [-1.5pt] 1.4 |
| Text only | 36.7 [-1.5pt] 11.5 | 0.0 [-1.5pt] 0.0 | 5.4 [-1.5pt] 1.7 | 23.3 [-1.5pt] 11.5 | 6.7 [-1.5pt] 5.8 | 5.0 [-1.5pt] 1.4 | 100.0 [-1.5pt] 0.0 | 0.0 [-1.5pt] 0.0 | 11.5 [-1.5pt] 0.3 | 3.3 [-1.5pt] 5.8 | 3.3 [-1.5pt] 5.8 | 100.0 [-1.5pt] 0.0 | 100.0 [-1.5pt] 0.0 | 97.0 [-1.5pt] 5.2 | 97.0 [-1.5pt] 5.2 | 60.1 [-1.5pt] 3.4 | 34.5 [-1.5pt] 1.0 | 7.3 [-1.5pt] 1.1 |
| Simple OA | 100.0 [-1.5pt] 0.0 | 0.0 [-1.5pt] 0.0 | 27.5 [-1.5pt] 0.0 | 40.0 [-1.5pt] 0.0 | 10.0 [-1.5pt] 0.0 | 6.7 [-1.5pt] 0.0 | 100.0 [-1.5pt] 0.0 | 0.0 [-1.5pt] 0.0 | 11.5 [-1.5pt] 0.0 | 100.0 [-1.5pt] 0.0 | 100.0 [-1.5pt] 0.0 | 100.0 [-1.5pt] 0.0 | 100.0 [-1.5pt] 0.0 | 100.0 [-1.5pt] 0.0 | 100.0 [-1.5pt] 0.0 | 90.0 [-1.5pt] 0.0 | 51.7 [-1.5pt] 0.0 | 15.2 [-1.5pt] 0.0 |
| VLM (ours) | 100.0 [-1.5pt] 0.0 | 80.0 [-1.5pt] 0.0 | 3.1 [-1.5pt] 1.3 | 70.0 [-1.5pt] 0.0 | 60.0 [-1.5pt] 0.0 | 2.5 [-1.5pt] 0.0 | 75.8 [-1.5pt] 5.2 | 57.6 [-1.5pt] 5.2 | 7.5 [-1.5pt] 1.3 | 100.0 [-1.5pt] 0.0 | 100.0 [-1.5pt] 0.0 | 100.0 [-1.5pt] 0.0 | 100.0 [-1.5pt] 0.0 | 90.9 [-1.5pt] 0.0 | 90.9 [-1.5pt] 0.0 | 89.4 [-1.5pt] 0.9 | 81.4 [-1.5pt] 0.9 | 4.4 [-1.5pt] 0.4 |
| Correct rejection | False rejection | ||
|---|---|---|---|
| Method | Non-recoverable hazard | Recoverable hazard | Safe variants |
| VLM + Task | 96/96 | 63/93 | 42/90 |
| VLM + Task + Scene Text | 96/96 | 83/93 | 58/90 |
| VLM + Task + Scene Image | 96/96 | 78/93 | 49/90 |
| VLM + Task + Scene Text + Tool Calling | 96/96 | 24/93 | 30/90 |
| Method | Laptop–Cup | Steak Dripping | ||||
|---|---|---|---|---|---|---|
| SR | SSR | Viol. | SR | SSR | Viol. | |
| No context | 100 | 20 | 80 | 100 | 20 | 80 |
| Simple OA | 100 | 20 | 80 | 100 | 20 | 80 |
| VLM (ours) | 100 | 100 | 0 | 80 | 80 | 0 |
Robot Safety
Robot safety studies how to prevent harmful physical actions while enabling robots to complete tasks around people and in uncertain environments.
See all 252 papers