Andrej Karpathy, Senior Director of Artificial Intelligence (AI) on the Autopilot team, recently presented at the Conference on Computer Vision and Pattern Recognition on Tesla's choice to remove radar from its perception system.

He echoed Elon's recent tweet: camera-based vision is more precise than radar, and radar contributes more noise than valuable data in a sensor fusion system.

Karpathy compared vision-only and sensor fusion performance when a vehicle approaches and then passes under a bridge, concluding that the vision-only version of FSD (Full Self Driving) had a superior vertical resolution, which helped distinguish the bridge from a stationary object that would cause the car to stop.

Why have most other firms, unlike Tesla, chosen to integrate data from low-cost radar and LiDAR in the autonomous solutions space?

One argument may be that Tesla's 1.3-million-unit vehicle fleet allows it to train vehicles using 6,000 hand-picked clips from a new release.

In other words, due to the size of its fleet, Tesla may have "won" the right to remove radar from Autopilot, which we feel is a substantial edge over its autonomous competitors.