Today's virtual worlds are not accurate enought to allow to develop perception algorithms in simulation. In order to develop sensor fusion, you also need to simulate the output of all sensors including their specific characteristics. Apart from the model quality, there is another challenge: Simulation runtime, which is substantial (!) and - to my best knowledge - not even close to realtime.
If you want to develop driving algorithms that sit on top of the perception stack, then this becomes simpler. You can work on the object level (object being simulated cars, pedestrians, ...) and statistically model perception errors. This is a lot faster, which is e.g. important, if you want to run large-scale reinforcement learning to develop your driving strategy.
In any case, in the future, I would say that we will see a lot more simulation (don't forget, all major players build heavily on simulation - just look into the numbers on how many miles Waymo simulates every day) and potentially going down all the way to the sensing level, because it allows you to develop and especially debug along the whole sense - plan - act stack.
Also, today there exist also hybrid approaches. You take real recordings and abstract them into a simulatable format that you can then, e.g., use to variate and derive artificial scenarios for simulation. This can be used to analyse the influence of different situative paramters on the behavior of a function to pinpoint which parameter(s) caused certain troublesome behavior that have been observed in real drives.
Hm, I would say that some consolidtion happened with that deal. Quote: "VW is also handing over Autonomous Intelligent Driving, the self-driving subsidiary that was launched just two years ago to develop autonomous vehicle technology for the Volkswagen Group. AID is valued at $1.6 billion.
The Munich-based AID team will become Argo’s European headquarters, a move that will expand its staff 40% to more than 700 employees."
AID has been Audi's L4/L5 development arm, which was separate from VW's main research arm. It also helps to judge the size of this deal, because VW only seems to invest one billion in capital and the remaining 1.6 billion investment is AID. It would be interesting to know, how they came up with that valuation ...
At least you will have autnomous traffic jam assistants much earlier, which allows you to spend more time on HN while stuck in traffic on your way to work on true self-driving vehicles ;)
It's best not to think in GB terms when talking about AD datasets. E.g., when you record raw data of a multisensor setup (lidars, radars, cameras), the data rate can reach 10+ TB/h. Camera-only datasets are in comparison much smaller.
Taken out of the argoverse dataset description:
- One dataset with 3D tracking annotations for 113 scenes
- One dataset with 327,793 interesting vehicle trajectories extracted from over 1000 driving hours
- Two high-definition (HD) maps with lane centerlines, traffic direction, ground height, and more
Technology: BMW's bet was on lightweight materials and less on battery technology as an enabler for ecological, electric vehicles. Production is still costly and more complex than for e.g. the 3 series. Currently, there is only one plant that is equipped with the necessary manufacturing capabilities (Leipzig).
Manufacturing: You need to adapt the assembly lines, which is not as trivial as it may sound. BMW's future platforms will allow assembly of EV's along ICE's on the same belts! At that point, you can produce EVs in all of BMW's plants (... with some modifications)
Financial: BMW is not really making money with the i3, production is too costly. Given that BMW is a public company, you don't want to mass produce a product that eats your profit. Try to explain that to "traditional" shareholders that like their dividends. TSLA is an entirely different stock, which can do what it does due to its growth story.
Also: Plants cost a lot of money build and operate. Stopping the production in a plant for a while to remodel it for a different assembly process as well (... not to mention the retraining of all the workers, new logistical processes). The magnitude is somewhere in the million dollars per day.
Human factors: In German labor law, you can't just fire people. Giving up on ICEs means that you have a lot of people that are not qualified for alternative technologies and need to be retrained. This does not happen in days or weeks - at the scale of BMW this takes years and of course is happening.
Sourcing: You need to source all the materials, incl. batterie. My guess is that BMW wanted to work with established players that know car manufacturing. However, that was probably a mistake given that e.g. BOSCH decided to quit (https://electrek.co/2018/02/28/bosch-gives-up-battery-cell-p...)
Shareholders: We had the topic before, but BMW is special here that the Quandt family owns roughly 46% of BMW. Accordingly they have a major influence on all decisions at BMW. They act definitely not like an Elon Musk - they are very cautious and strategic. They also seem to act labor friendly and have a positive impact on the work conditions at BMW (... again meaning that they will make sure that BMW doesn't have to let go a lot of people).