Trucks and trailers can now estimate depth from cameras alone, without expensive sensors

A new method lets articulated vehicles (trucks, trailers, robotic platforms) estimate 3D depth using only cameras, without LiDAR or other sensors. This matters because it's cheaper than current approaches and works across different vehicle types, which could eventually make autonomous driving systems more affordable.

Until now, depth estimation methods for autonomous vehicles assumed a single rigid body. Trucks bend — the trailer moves independently from the cab — which broke those methods. This paper solves that by using the constraint that the truck's geometry has to stay consistent across multiple camera views, even as parts move relative to each other. If this scales, fleets don't need to buy expensive sensor packages to perceive depth; cameras alone might be enough. That's a cost problem solved.