Tesla Motors hiring talent for developing self-driving cars
efinancehub.com
efinancehub.com
ADAS isn't just about the self-driving cars that Google is testing. ADAS includes systems like:
- Warnings if you drift out of your lane (Mercedes has been advertising this for four years)
- Warnings and stops if the car predicts an accident may occur soon (Infiniti has been advertising this for years)
- Automatic parking (Ford has been advertising this for years)
- Blind spot detection (many car companies have it)
- Cruise control that maintains minimum distance to the car in front of you (BMW and Audi mid and high-end cars had it since 2010)
It really looks like Tesla is trying to develop those features for the Model S and future cars.
Then a step up on the "drifting out of your lane" is one that will attempt to keep you in your lane (Acura and Mercedes have this, goes by the name "Lane Keep Assist"). [0]
Seems saner than Google's approach, but Google doesn't manufacture cars so the same model would be harder...
And for a company like Tesla, who will sell their own cars, an incremental approach can be self-funding. Google has a ton of money to dump into research and is unlikely to bring a car to market themselves, so building a finished product will make it easier to find customers to license individual components of their tech to.
I'm sure the NSA is working on something right now...
I was wondering what data they were gathering, and what they would do with it, and how it could be useful to other drivers:
What if you have a Tesla driver profile and you could see your driving habits ranked amongst all other tesla drivers (anonymously) to determine your rank as the most safe, efficient, slow, fast, whatever driver.
To see where you stack among others would have an effect on your driving. To see how your particular car performs against all others - its mileage vs age against others etc...
I think that there is a whole other layer that could be added onto the Tesla experience that other car manufacturers are far too old and less nimble to do appropriately.
It was also reported by Wired the same day. http://www.wired.com/autopia/2013/09/tesla-autonomous/
He's also working on automated maintenance facilities that would allow a self-driving car to report for maintenance/cleaning. The car drives up, receives service and drives off afterwards to the next pick-up.
The "automated fleet" idea is not how people are planning to see this idea play out. Automated, self-driving cars are one thing, but an automated fleet makes it sound like a train rather than many vehicles operating very efficiently (meaning without human drivers) but each with a separate destination.
To an alien watching from orbit, a country of automated cars would appear different from the present only in that the cars would be much closer together than they are now, with no increased danger of collision -- indeed, less danger. Cars would still leave the freeway and go to their intended destination just as they do now, just more efficiently.
> it wouldn’t make economic sense to use a self-driving car service to get to and from work.
It would if everyone still owned a car as they do now, the only difference being that the car drives itself. If the idea of a "fleet" were taken seriously, why not switch to trains or buses instead of cars? The only reason to have a car is to be able to go directly to your destination.
If only ¼ of car owners are on the road at peak traffic times then it could make sense to have automated car rental for a specific trip instead of car ownership. Assume drives to work take 30 minutes and that most people go to work from 6:00 AM to 9:00 AM. Also assume that we have inefficient routs requiring the car to drive the same distance for all pickups and drop-offs. Between 5:30 AM and 8:30 AM you could get three people to work with one automated car. Assuming this would it be cost efficient for companies to charge a low enough trip cost to make it more economical for people with fairly predictable driving requirements to rent a car as needed instead of buying one.
We wouldn't need to dedicate such a high level of inefficient resources to parking vehicles that sit mostly unused and it could offer us the opportunity to remake our cities to be more pedestrian and bike friendly.
When I hear the word "fleet" it usually just means a number of vehicles bought by one company: A car rental company has a fleet of cars. A trucking company has a fleet of trucks. A car dealer has a fleet manager. (Forgive me for pointing this out, I'm sure you know it already, just trying to clarify the terminology.)
Enterprise Rent-a-Car will already send any of the cars in their fleet to your door. Of course they send a human along to drive the car to you.
If they got an automated fleet they could offer the same service they do now, but they would just send the car over by itself.
At least that's what "automated fleet" sounds like to me. :-)
Both of which are kind of a big deal.
My car isn't just a train or a bus that goes directly to where I want it to; that's called a taxi. My car is a small extension of my house.
However I will say that your issues are pretty easily solved, imo. The car service simply customizes your ride experience with your radio choices. Car seats are necessary because of imperfect human drivers and (?) not necessary when robots are driving.
Removable/secure trunks could be stored at some site until needed, retrieved from cars and stored by robots since lifting and carrying are simple tasks for robots. When you request the next leg of your service, the first stop is the storage center where your secure trunk was stored.
I already mentioned about the maintenance facilities Musk is working on. Cheap and effective sensors detect gum/soilage, modern fabrics repel dirt and detritus and are cleaned to tolerable levels. Or another car comes instead. And your 2 hours in a car drops precipitously as smart-queuing greatly improves travel times.
The car service makes car pooling and mass transit more practical as well. Maybe you could hitch a ride with someone to work, but then you might have to stay late. With personal car ownership, you need to drive alone both ways, but with the car service, you can share the morning trip and if you need to take separate transportation in the evening. If buses in your area can't reliably get to work on time, you could have the car service pick you up to go to work, the the bus take you home.
And something similar on Singapore's roads too: tolls vary by time of day. http://www.lta.gov.sg/content/ltaweb/en/roads-and-motoring/m...
Finally, taxis have a pricing feature common in many places: a surcharge for certain times of day (can be peak times, can be nighttime...).
But that's just me. The normal price to get to work on the MRT is less than 1 USD. So the difference between that and free isn't worth it for me to alter my schedule. Less well off people might be more influenced, and raising prices might be politically infeasible. (And I usually ride my bicycle to work anyway.)
Also, it's not a mass transit system, but the 520 Toll Bridge in Seattle uses this to price tolls. It was the first time I'd seen this, but perhaps there are other places where tolls operate similarly?
What are the energy cost implications of this? Okay, I drive my car to work, and it sits there until I go home. If it was part of a fleet, it would leave my work parking lot, go to some other part of town (empty, using energy), and so on. Okay, we can also save energy by ride sharing, but I speculate that it wouldn't take much to make that really unappealing. In reality, we don't just go from home to work and back; we stop at the grocery, drop the kids off at school, and so on.
We already have fleets of automatic cars - they are called taxis. Their neural net is wet, is all. So as a first approximation I'd say taxi fleets are a pretty good proxy for energy use and cost. Sure, the driver needs to be paid, but we can model that.
In your case I know it's not as applicable, but I often recommend to people that pay off their car loans that they should continue making that payment, just into a savings account. In less time than it took to pay off the loan (remember, you were paying interest too), you'll have saved enough to pay cash for your next auto. If you've been able to live with that money dedicated towards a loan payment, hopefully your budget can continue not to depend on the "extra" resulting from not having the obligation.
It might turn out that by the point the technology exists for self-driving cars, private car ownership will be extinct and driving a car within city limits a felony.
Tesla is operating in the upper luxury end of the market, where margins are fatter, where vehicles sold can be lower but the margins maintained. They are making very large cars, even their follow up model to the S is the same wheel base. Why is that? Most likely because Musk knows they cannot deliver the range needed at cost and more importantly, a size, that a smaller car will be.
There are many self driving efforts going on in the market, both Mercedes and Infiniti have vehicles on the market that offer some limited form of self driving. Mostly lane departure and the like but they are pushing distances.
Tesla should be the one who worries most, because they have a unique product only for a short time and they won't get into the smaller sized car market because the technology path they have followed is not progressing fast enough.
If they made a car the size of standard midsized car, lets pick Fusion or Accord, the wheelbase shrinks by ten inches on average, width comes down too. All that shrinkage will take away from battery capacity, hence range. That is one reason many of the offerings by the incumbents have less range. Many use a battery solution not all that different, just restrained by size and technology.
There was a reason Musk went for the price point he did, it allowed larger vehicles which current technology requires. Never bet on the incumbents sitting still, Musk certainly is not. It just comes down to, can Tesla execute well in the mid size market, should they ever get there. Who knows, he made cede that space by licensing tech to let other people work the small margin sales
Musk makes no secret of his positioning in the market and going for the price point that they have. But there are hundreds of different cars at the same price point of the Model S, none of which are even in the same ball park.
The necessity for the incumbents to appeal to the mass-market is exactly why they should be worried, they're business relies on so many manufacturing synergies to reduce price, adapting all of those chassis and systems to cope with electric drive and self driving is going to be a challenge if they're to sell to their exist mass-market customers, at a price they expect. Tesla doesn't have that problem.
Tesla is delivering 550 cars per week[1], which is about 2357 per month. In August Nissan sold 2420 Leafs[2]. So seems like sales are just about equivalent. Except that Tesla is supply constrained and sells direct to consumers, so that 2357 underestimates actual sales. And many of those Leafs are sitting in dealerships, while none of the Teslas are.
Now Telsa only sold 5k cars in 2012, so it's doubtful they have on the road more than half of the 75,000 Leafs that have been sold[3]. But the Leaf started shipping in December 2010[4] whereas the Model S started shipping in June 2012[5] so it's not exactly a fair comparison.
The Chevy Volt actually did better than either in August, with 3351 sales[2]. Although their year-to-date sales were the same as the Leaf (14k) which suggests they had some late resurgence.
[1] http://insideevs.com/elon-musk-talks-all-things-europe-and-u...
[2] http://green.autoblog.com/2013/09/04/chevy-volt-nissan-leaf-...
[3] http://green.autoblog.com/2013/08/21/nissan-leaf-production-...
At this point one can only talk about installed base as both Nissan and Tesla cannot meet demand. Nissan makes its own batteries but cannot make them fast enough. And of course the markets are fairly different. The LEAF is much, much cheaper, and is in no way a luxury or performance car.
A lot of the difference between the (for want of a better word) "software" mentality and the "hardware" mentality comes from how we approach risk. One approach emphasises mitigating the impact of adverse events by emphasising flexibility and agility; the other approach emphasises minimising the chances that adverse events occur at all, by emphasising predictability and control.
In other words, do you design so you can fix it quickly when it breaks (at the expense of having it break often), or do you design it so it very rarely breaks (but when it does, it is more expensive to fix)?
The answer to this question not only depends on how safety-critical the system is, but how complex it is too. The prediction-and-control approach rapidly becomes untenable when you have systems that reach a certain level of complexity -- the cost of accurately predicting when failures are going to occur rapidly becomes larger than the cost of the failure itself. As the complexity of the system under development increases, the activity looks less like development and more like research. The predictability of disciplined engineering falls apart in the face of sufficient complexity. Worse; complexity increases in a combinatorial manner, so we can very easily move from a predictable system to an unpredictable one with the addition of only a small number of innocuous looking components.
Most forms of mechanical engineering emphasise the second (predictive) approach, particularly for safety critical equipment, since the systems are simple (compared to a lot of software) and the costs of failure are high. On the other hand, a lot of software development emphasises the first (agile/reactive) approach, because the costs associated with failures are (normally) a lot less than the costs associated with development.
Of course, a lot of pejorative terms get mixed up in this, like: "sloppy engineering", and "cowboy developers", vs "expensive failures" and "moribund bureaucracy" but really, these approaches are just the result of the same cost/benefit analysis producing different answers given different input conditions.
Problems mainly arise when you use the wrong risk-management approach for the wrong application; or for the wrong part of the application. Things can get quite subtle quite quickly.
One of the challenges in developing automotive ADAS systems is that a lot of the software is safety critical, and therefore very expensive to write, because of all of the (necessary) bureaucratic support that the OEMs require for traceability and accountability.
Equally, a lot of the advanced functionality for machine vision / Radar / Lidar signal processing is very advanced, and (unfortunately) has a lot of necessary complexity. As a result it is very very costly to develop when using the former approach; yet may be involved in safety-critical functions.
This is not by any means a solved problem, and very much requires detailed management on a case-by-case basis.
Certainly testing infrastructure becomes much more important as the sensor systems that we develop become more complex. (Disclaimer: my area of interest). Indeed, my experience indicates that for sophisticated sensor systems well over 80% of the effort (measured by both in hours of development & size of the code-base) is associated with test infrastructure, and less than 20% with the software that ends up in the vehicle.
Perhaps the word "test" is a misnomer here; since the role of this infrastructure is not so much to do V&V on the completed system, but to help to develop the system's requirements -- to do the "Data Science" and analytics that are needed to understand the operating environment well enough that you can correctly specify the behaviour of the application.
http://spectrum.ieee.org/automaton/robotics/artificial-intel...
Videos here:
There might not even be windows in the self-driving car of the future.
"The three point belt was developed by Nils Bohlin who had earlier also worked on ejection seats at Saab.[15] Volvo then made the new seat belt design patent open in the interest of safety and made it available to other car manufacturers for free.[16][17]"
It's the same as saying one phones corners can't be perfectly round or icons can be on a certain side of the screen.
"They will have to be proven as perfectly safe. One accident and people will go crazy."
And though that's true, it of course made me think about the countless accidents involving human-driven cars through distraction, drink/drugs, weather, medical issues, confusion and whatever else.
We haven't proven our driving abilities as perfectly safe, not by a long shot!
There is even a disruption potential: Put together a new insurance company that only insures self driving cars and make it meaningfully cheaper. The ones that still insure self directed cars and self driving cars both will have a tough time to compete because their blended economics don't allow for the super cheap rates.
I would bet the headlines and maybe a politician or two would yell it, but I would also guess that the overall reaction would be much more reasonable.
I'm keen to see self-driving cars, but there will be a lot of opposition. Lobbying from trucker unions and the like will push negative stories and pressure decision makers, for example.
They've produced several different VAG-based models including several Audis that mainly focus on self-driving and similar AI. One of their more high profile recent examples was an Audi TT that climbed Pikes Peak without any human input. [3]
Some pretty cool stuff, and likely the cars you saw.
[1] http://www.stanford.edu/group/vail/ [2] http://en.wikipedia.org/wiki/DARPA_Grand_Challenge [3] http://www.youtube.com/watch?feature=player_detailpage&v=Y88...
I definitely agree that more competition in the area is good. Hopefully as more players build these systems, they'll continue to improve and be commercialized quicker.
So, I don't think it lacks competition for Google, who doesn't even build their cars - most probably they'll license the technology for other companies, and Tesla may well be one of them.
[1] Links from 8 to 14 in this Wikipedia page: http://en.wikipedia.org/wiki/Autonomous_car#cite_note-idsia-...
[2] http://phys.org/news/2013-03-japan-group-fuel-saving-driverl...
I'd be willing to bet that they were planning some sort of partnership. This isn't a technology that you want to get wrong, and once other vehicle manufacturers start offering a comparable electric car (soon) then Tesla will need to have something to differentiate themselves.