Computer Vision & Perception
RGBD depth cameras, lidar and ultrasonic layers fuse into one occupancy picture — people, chair legs, glass and 5 cm thresholds, read at up to 30 m.
A fused occupancy picture
RGBD depth cameras, lidar and ultrasonic sensors publish into one model of the world: people, chair legs, glass doors, 5 cm thresholds and 30 m of forward radar coverage. No single sensor is trusted alone — glass reads in ultrasound, dark floors read in vision, crowds read in lidar.
Human intent, not just human position
Depth cameras classify motion, not just distance: a person walking toward the robot reads differently from a person standing in conversation. Delivery robots slow, announce with light and voice, and re-plan — instead of braking mid-corridor.
The same stack in every chassis
Perception is one platform across D1, C1, C2 Pro and G1. A lesson learned on a scrubber's night program ships to every fleet in the next release — which is why mixed fleets get safer, not more complicated, over time.
Talk to an engineer
Integration questions answered by the people who ship the stack.
