We've experimented with it a bit this last summer - while there's progress, there is no expected date. But it's something i'm working on in "Spare" time

Thanks for the reply. Just curious if it was still on your list.

Ron

You bet not much leaves the list. Priorities move things around to make development across projects a smoother transition. We had expiremented with using light beacons and infrared beacons on the robot. The concept will be something like that...

The light beacons had trouble in varying lighting conditions.

The infrared is an entirely new camera and therefore doesn't view the real world view so you have nothing to compare to where the robot actually is. We would need to consider a dual camera (ir and regular) unit for it to work correctly.

Perhaps tapping into the "google cloud api" could help, just for identifying the item to avoid?

There's object recognition in ARC, which is faster than having to transmit data over the internet and wait a minimum of 500ms for a response. That's not how it works anyway - the camera is mounted in the corner of a room looking down at the room. The robot has a beacon which the camera picks up. You highlight areas that the robot must avoid and highlight paths to get into other rooms.

It's how swarm robotics work, such as the dancing drones. They have a large number of IR cameras (sometimes up to 50 cameras). The IR cameras do not need to see objects to avoid, so that part is easy. Mixing the two is the challenge, but it's doable with a dual camera PCB as i mentioned.