All projects
Status: Conceptresearch · Concept
AI Vision for Rescue Robotics
A research direction, not a build: applying computer vision to the disaster robot so it can distinguish a person from surrounding debris and assess whether a path is passable.
Concept: Designed on paper. Not built yet.
Placeholder artwork. Project photography, screenshots and video are pending — see the asset checklist in the repository.
Why
Thermal and life-detection sensors report that something is there. Vision could report what it is — reducing false positives and helping responders prioritise.
Open questions
This is unbuilt work. The questions below are what would need answering first.
- What training data exists for partially occluded people in rubble?
- Can inference run on the robot's power and compute budget?
- How is a false negative handled when a life is at stake?
Possible future improvements
Status: ConceptNone of the items below have been built. They are recorded here as direction, not capability.
- Dataset collection from staged rubble environments
- On-device inference benchmarking
- Sensor fusion between vision and thermal channels