I’m not aware of any HD maps built entirely automatically. As soon as manual steps creep in, the cost skyrockets. The backhaul and processing power required for updates can also be staggering.
The wages paid to professional drivers is also staggering.
If amortized cost of hardware + backhaul & processing costs < professional driver wages, then the model works. This is especially true as tech improves and wages increase. There will be a crossover eventually.
Back of the envelope calculation for HD mapping cost: $5k / km * 4,000,000 km of road in the US = $20bn / yr on HD mapping (assuming once a year updates, which is actually quite infrequent). And that’s not considering all other costs involved such as $500k in sensors per vehicle, safety techs (teleoperations are still super common), etc. It has to come down in cost a lot to compete with the price of unskilled labor.
$500k in sensors per vehicle is off by a substantial margin -- nearly an order of magnitude. I also seriously question your $5k / km "napkin math" without any elaboration.
I would also point out that there are 2M truck drivers earning $47k per year [1]. That's $94B annually in pure labor cost, and doesn't even account for rideshare & taxis.
Sure, but the point is that the car can flag something that doesn't look like the map so a human can review it. ie. closed road, new road, construction markers, etc.
That's way less intensive than having to build the whole thing from scratch or having it be fully automated.
Humans are good at exceptions. Let the humans do that while the machines do the rote task of "Yep, road still looks like the map. Nothing to see here."
I think there’s way more hand correction needed than people are estimating. Google Maps (the non hd version) employs a small army of 6k contractors [0] to correct way simpler data. Getting cm accurate lane markings and correctly classifying permanent and transient features of a scene is extremely error prone and I would guess at least an order of magnitude more manpower to make this happen at scale.
Why do you think the cost of building the HD maps is prohibitive? How much do you think it would cost to say, keep a square mile of a dense urban HD map updated for a year? It's probably way lower than you think.
I didn’t say it was prohibitive but it’s certainly not cheap. The cost is per km rather than per square km. DeepMap charges $5k / km for example (the big companies don’t publish their costs, but it’s widely known in the industry to be on par with this). I can’t seem to find the stats again, but map invalidation happens on the order of weeks to months.
