TrisTrash: the street becomes a sensor
Every truck already driving through the city can look at the sidewalk. TrisTrash turns the cleaning and collection fleet into a sensor network: onboard camera, edge inference, geolocated alert and closed-loop dispatch.
The problem
Fixed routes miss whatever happens between passes. A bin overflows at 11:00 and the service doesn’t come back until the next day. Abandoned waste waits for someone to report it. The citizen gets there before the report does. That’s what gets tendered, fined and tweeted about.
How it works
Onboard camera on vehicles that are already on route. Edge inference (no need to send video to the cloud to decide; reference field stack: NVIDIA Jetson Orin Nano, DeepStream and YOLO). Detection of overflows, abandoned waste and other visibly operational incidents. Geolocation of the event and entry into the workflow (GIS / operations). Assignment to the nearest unit and closing of the loop. In Madrid, the removal target for alerts covered by the system is under 4 hours, 24/7.
What we’ve measured
Who it’s for
City councils and cleaning/collection contracts that need evidence, not a slide deck. Also internal operations teams tired of flying blind between routes.