Just imagine Tesla would have subventioned passive LIDAR on every car they ship to collect data. Wow that dataset would be crazy, and would even improve their vision models by having ground truth to train on. He’s such a moron
All of that combined is probably closer to $1k than to $140.
And, again, that's - what - 10 years after Tesla originally made the decision to go vision only.
It wasn't a terrible idea at the time, but they should've pivoted at some point.
They could've had a massive lead in data if they pivoted as late as 3 years ago, when the total cost would probably be under $2.5k, and that could've led to a positive feedback loop, cause they'd probably have a system better than Waymo by now.
Instead, they've got a pile of garbage, and no path to improve it substantially.
They might be!
But I doubt it.
I don't know enough about Tesla's cameras, but it's not implausible to think there are LIDARs of low enough quality that you'd be better off with a good quality camera for your sensor.
Again, I doubt this is the case with BYDs cameras.
But it's still worth pointing out, I think.
My point is, BYD's LIDAR system costing $x is only one small part of the conversation.
>The EX90's LiDAR enhances ADAS features like collision mitigation and lane-keeping, which are active and assisting drivers. However, full autonomy (Level 3) is not yet available, as the Ride Pilot feature is still under development and not activated.
Solid-state LIDAR is still a fairly new thing. LIDAR sensors were big, clunky, and expensive back when Tesla started their Autopilot/FSD program.
I googled a bit and found a DFR1030 solid-state LIDAR unit for 267 DKK (for one). It has a field of view of 108 degrees and an angular resolution of 0.6 degrees. It has an angle error of 3 degrees and a max distance of 300mm. It can run at 7.5-28 Hz.
Clearly fine for a floor-cleaning robot or a toy. Clearly not good enough for a car (which would need several of them).