Tesla on Autopilot

Tesla on Autopilot

Ultimately, driverless cars are the future. However, if my MS drove itself, I wouldn't have paid the premium for the performance it has. Possibly this is being considered for the Model E.

Andrzej1 | 18 September, 2013


“However, if my MS drove itself, I wouldn't have paid the premium for the performance it has.”

Why not?

Driverless cars are your best hope of ever being able to have your vehicle legally travel at its engineering limits tempered only by the limits of human biology.

TikiMan | 18 September, 2013

IMHO, automated vehicles are the only solution to traffic gridlock on freeway systems in heavy populated areas. As it stands, building out a larger travle infrastructure is a never-ending battle, as by the time roads and freeways are improved for better traffic flow, population growth has already caught up with the improvements (thus making the improvements moot).

The only logical way to improve freeway traffic flow, would be to decrease (or remove) the freewill of its controller (the driver/pilot). Thus, if vehicles acted as a single automated entity, all in direct comunication with each other for the good of the entire entity, traffic would always flow, and gridlock would be eliminated (as would driver error, and millions of high-speed accidents).

Unfortunately, humans prefer freewill, and like to have control of their vehicles (even if it causes harm to the overall good of the majority). Thus, I suspect HOV lanes would become future 'automated vehicle only' lanes, as commuters with automated vehicles would relax, work, or enjoy some entertainment, while their automated vehicles transport them from point-A to point-B. As more vehicles asimilate to automation, more lanes would become 'automated only' lanes. However, non-freeway roads, city / county streets, etc would be left to the freewill of its pilot, to choose automation or not (thus allowing for some freewill).

Brian H | 19 September, 2013

To judge by LA, the bigger freeways get, and the more numerous, the worse the traffic becomes. Extrapolating in either direction is problematic. ;)

grega | 19 September, 2013

Thanks for the link.

If the cars can become 100% driverless the implications are ... mind boggling really.
I know Elon has expressed a preference for assisted driving, and in this article he describes it as being far easier to get 90% there, which certainly makes sense and improves safety and traffic management.

At 100% the game changes.
* No more home garages in busy areas
* shopping centres designed with huge drop off zones and shared parking areas
* Auto charging.
* My car picks me up for work in the morning
* Only 1 car for me and my wife, it can return to her after dropping me to work.
* Back seats become the main seats, with front seats facing backwards with the kids?

Why have a car at all if I can instantly rent a car and have it turn up ready for my need, and when I'm finished it can go and charge itself while getting a quick vacuum, and head off to the next renter. Battery worries are non-existent if you rent for a specified trip - a car capable of that distance simply turns up and drives you there, and if it did break down another car would come by to complete the trip. Anybody who doesn't travel in peak could easily find a car.

Of course it goes much further than that. I see no reason I can't book a car trip from home to work, and have an electric car with 4 'business class' seats turn up, that I share with 3 others. Really a self-drive taxi, but with smart passenger management. Perhaps a seat in this car which drives me to a hyperloop station, where I switch into an identical seat to finish my trip.

Ah well, 90% of auto driving is a start. It doesn't change the industry, but it'll be a good step.

PBEndo | 19 September, 2013

@Andrzej1 I am totally in favor of moving towards driverless cars. However, the attraction of the acceleration and handling go out the window when a computer is driving. A high top speed is nice, but only if the regulations are changed to allow it and the computer is deemed safe enough to drive at higher speeds. I think it will be a long time before that happens.

Elon's plan to automate 90% of the driving concerns me. The transition from 0% automation to 100% may be a rough road. It has already been shown that minimal automation (cruise control) significantly lowers alertness in drivers. Imagine if 90% of the driving is done by the car. People will not be alert to handle the 10% that the computers can't do yet. I would love some clarification on what tasks the remaining 10% includes.

cfOH | 19 September, 2013

@pbendo I suspect the remaining 10% consists of complex navigation with short distances in places that computer vision systems won't really work well in. For example, pulling into a parking lot where you have to negotiate around a low wall, through a parking gate arm (which the car probably won't automatically pull close enough to by itself to let you operate the ticket kiosk), and then around the maze of cars to find a spot (it could do it, but it'd probably take far longer than a human to do so). Or backing out of a driveway into a street with traffic. In a nutshell, I think the beginning and the end of the journey is probably the hardest for a robot -- the 90% in between is probably what Elon is envisioning gets automated most easily.

PBEndo | 19 September, 2013

@cfOH I have heard others working on driverless cars say the most difficult thing is reacting to unexpected objects in the road.
If a human sees a cardboard box in the road, they can usually quickly determine if they need to stop, swerve, or if it is just safer to hit the cardboard. Getting a computer to make the right decision is quite a challenge. If Elon is referring to things of this nature, having the driver stay alert for the rare unexpected problems would be impossible.

Vicelike | 19 September, 2013

Fun Thread!

PBEndo | 19 September, 2013

"Thrun tries to put things in perspective. His self-driving cars, he says, don't make careless mistakes. The cameras never ignore a red light, and the radar reliably prevents rear-end collisions. "In those areas," he says, "robotic vehicles are already better than humans."

The driverless vehicles are worse, though, at reliably identifying objects. "That's something we humans are incredibly good at," Thrun explains. He picks up objects from the conference table in front of him to illustrate his point: "Here, a telephone. A roll of tape. It's not something we have to think hard about."

The cars' cameras see these things just as clearly as the human eye does, but the computers take longer to assess whether or not it would be dangerous to drive over them. What this means is that a robotic car will slam on the brakes even when the object in question turns out to be just a cardboard box blowing down the street, because it can't immediately assess whether the box isn't actually a baby carriage."

from here -

Zebuf | 19 September, 2013

On the other hand, a fully automated vehicle could have both infrared and ultrasound detectors, and make a very good guess at what was inside the box.

It would also communicate directly with all neighbouring vehicles, getting info on all relevant movements in advance.

With the processing speed of tomorrows computers, this would enable bumper to bumper traffic even at very high speeds, increasing the possible throughput on any highway.

It would be a very safe scenario, apart from that single, manually operated car that inevitably would f*** it up ;-)

PBEndo | 19 September, 2013

Would an autonomous car be intelligent enough to stop when the doors of the armored car in front of you open and the cash is falling out? Or would it assume it is just harmless paper and keep driving right past the cash?

mvannah | 19 September, 2013

Driving long distances could become like taking a red eye flight. You do your best to get some sleep while your car does most of the monotonous driving. The car could simply wake you up when it is time to recharge, encounters a construction zone, encounters an unknown gate, toll road, etc. Google maps already identifies road closures and construction zones, so the car could wake you up five minutes before you need to take the wheel again.

I only hope that they automate the home gate and garage door openers and allow the car to drive home from the airport by itself. I hate taking taxis and getting friends to drive me to the airport. I suppose inductive charging would be good too for the times that the car drives itself home and can't be manually plugged in.

An self valet function would also be awesome. They could set up a vehicle/parking structure communication system so that the car could be directed to the most efficient parking spot. I have already seen occupancy sensor lights at the PDX airport that identify individual parking spots that are occupied or vacant.

Convoy type drafting would also be a feature to improve efficiency. There would be pelotons of autonomous vehicles driving together, much like you see in the Tour de France where they take turns drafting.

Blind and other driving-impaired people will have more freedom to travel. I'm sure there are hundreds of other opportunities that will present themselves as well.

Andrzej1 | 19 September, 2013


Once Level 4 cars hit the road and human-driven vehicles are declared a public menace and banned from public roads shortly afterwards then the raison d'être for speed limits would vanish as well at least on highways and arterial roads.

Yes, I too have read the French study and wholeheartedly agree with you that this 10% business would be quite dangerous and ultimately untenable in any significant moving situation. Another example would be Air France 447 in which the autopilot automatically disengaged when the Pitot tubes froze over. The pilots who had lost situational awareness, panicked and drove an aeronautically sound airplane into the ocean killing all aboard. I would expect similar results on our highways.

Andrzej1 | 19 September, 2013


I totally agree with you in regards to driverless cars being a solution to gridlock. For the system to be truly robust you would also need V2I to implement demand-based pricing or some other allocation scheme in the event vehicular demand exceeded road capacity. However this would be very rare. Some advocates such as Thomas Frey claim a 10-20 times increase in road capacity with the advent of the driverless car and lane, distance, and time compression. In fact he actually misses one and that is demand-based dynamic allocation of lane direction. See here, .

Now let's examine the cost of human free will and many an individual’s strong desire to express it by driving one's car.

In the US, there are nearly as many roadway fatalities each month as occurred during 9/11. For the year 2000, NHTSA estimated the economic cost of motor vehicle accidents was $230 billion in the US alone. As for the annual global death toll due to car accidents, it is currently at about 1.2 million and estimated to reach close to 2 million by the year 2020 by the World Health Organization, if no further mitigation action is taken. Finally one has the congestion cost which in 2011, The Annual Mobility Report estimated at being over $100 billion for the US alone.

Given the above, I am confident that once Level 4 cars are roaming our public roads then human-driven vehicles will be phased out and ultimately banned from our public roads. We will continue to be able to express our free will by driving on private roads and tracks.

Andrzej1 | 19 September, 2013


“The cameras never ignore a red light”

Red light? What red light? See here,

carlk | 19 September, 2013

The best situation is I still can enjoy driving my car when I want to but I don't need to worry about driving when I'm tired and not wanting to or if it's even dangerous to do so.

JZ13 | 19 September, 2013

Why do so many people view autopilot as an either/or? It would obviously have an "OFF" button so that you could take control of the car whenever you want. Enthusiasts can drive their performance to their hearts content and then turn on the auotpilot when they commute to work so they can multitask.

I think autopilot will increase restaurant revenue as more diners opt to have that extra glass of wine knowing the car will drive them home. Although I'm sure DUI laws will not change for a very long time. I can see some folks who are over the limit getting caught by DUI checkpoints and some others by getting into an accident caused by another driver.

PBEndo | 19 September, 2013

@cfOH I don't think the parking maneuvers are the problem. Many cars have already been on the road for years with parking assist features.

TikiMan | 19 September, 2013


I agree!

bb0tin | 20 September, 2013

Autonomous cars will reduce gridlock for another reason: you can share a car with other people without effort. You book the trip you want to make, the system analyses all the other cars and booked trips, and diverts one to pick you up. Instead of one person per car, on average, during peak hour, you could have 4 or 5. It would be faster and cheaper for each passenger since they do not need to park and walk, and share the cost with the other passengers.

Andre-nl | 20 September, 2013

"Enthusiasts can drive their performance to their hearts content"

A nice eufemism for 'street racing'? ;)

As soon as autopiloted cars are safer than humans, tolerance for human error will take a nosedive, and 'driving to your hearts content' will become increasingly difficult to practice.

I expect in the not so far future (eg 20-40 years) that driving a car on public roads will be outlawed and you can only self-drive your car on a racing track. Maybe you can self-drive with the autopilot in 'nanny mode', preventing you from making fatal mistakes.

Andre-nl | 20 September, 2013

Btw, there was a nice article on Green Car Congress about how far Mercedes is with self driving cars. This technology seems to be moving fast and Elon doesn't seem to me to be the person who takes a wait-and-see approach.

PBEndo | 20 September, 2013

@Andre-nl "I expect in the not so far future (eg 20-40 years) that driving a car on public roads will be outlawed"

Once driverless cars are accepted and proven safer, that would make sense in terms of safety and efficiency. However, I expect it will be a long time after the technology is ready before cars that aren't completely autonomous are banned from the roads. That type of legislation usually faces a lot of opposition. It would suddenly make any older cars worthless. What politician wants to be responsible for making countless cars worthless? Typically, exceptions/loopholes exist for a very long time after new technology is implemented and required on new cars that allows the old cars to remain in use.
Think of how many collectors cars or just plain old cars you see on the roads that don't have airbags, anti-lock brakes, cyclops brake light, etc.

Unfortunately, I think this will delay implementation of some of the major benefits of autonomous vehicles. They won't be able to operate as efficiently or as safely when there are still humans controlling some of the cars on the roads. Initially, it will probably work like the HOV/Carpool lanes do today. Dedicated lanes will allow driverless cars only, where they can travel at higher speed and in much closer following distances.

avanti | 20 September, 2013

There is an amazing book called "Magic Motorways", written in 1940 by Normal Bel Geddes, based on the research he did in designing the GM Futurama ride for the 1939-1940 Worlds Fair. It anticipates most of the ideas discussed in this thread, along with many others. My favorite is his observation that if you had a fully-automated highway system, highway crossings could all be at grade, with the vehicles interleaved at the crossing via proper timing! What an image.

You can read the book on-line for free at:

jbunn | 20 September, 2013

Traffic on the highways in San Francisco is epic, but the highways terminate into a city of unsynchronized traffic lights on very short blocks. Traffic in-city moves perhaps 7 miles an hour or less. No matter how well the platooned cars move on the major roads, they eventually terminate into a slowly moving parking lot.

This may be unique to us, but in our case the plumbing is stopped up at the ends, not in the middle. A sane solution would be to synchronize lights to allow traffic to move more smoothly.

Andrzej1 | 20 September, 2013


“I expect it will be a long time after the technology is ready before cars that aren't completely autonomous are banned from the roads.”

Once Level 4 cars hit the road the transition will be pretty quick. The cost of maintaining the status quo in terms of human life, injury (WHO estimates 50 million people injured annually in motor vehicle accidents) and suffering, and coin is simply too high. The US is spending trillions to avoid another 9/11 and subjecting us all to the indignity of being molested at airports in order to prevent another 9/11 yet a 9/11 event happens every month on US highways. Many roads are being used at 1/10th or even 1/20th their potential capacity and we are forced to spend tens of billions on rail links to combat congestion. The $94.1 billion California High-Speed Rail project would never be built if level 4 cars were available today.

Lets take your position and assume there is no mandated transition or is greatly delayed. Regardless, many millions of Level 4 cars will appear on the roads. More and more the focus will inevitably shift to the loss of life caused by human-driven cars. After each successive fatal accident, headlines will scream ever louder, “When will this carnage end?”. Public opinion will quickly shift and human driving will rapidly become socially unacceptable and will deemed as reckless behaviour. If the government does not act, then the lawsuits will begin. Once the lawsuits begin, watch the insurance companies flee. At this point the ban is inevitable.

Will people be outraged? Not really since at this point it will have been demonstrated how much superior computer driven vehicles are and how our lives can be enriched by their use. Imagine going to sleep in your car and waking up in your destination city a thousand miles away or the elderly who would never have to worry about losing their driver’s license.

“What politician wants to be responsible for making countless cars worthless?”

Not necessarily worthless. Perhaps you will be able to convert your existing Model S to driverless operation. See here, , and here, , and here,

Andrzej1 | 20 September, 2013


Thanks for the link!

The modern version is called Autonomous Intersection Management(AIM). A short simulation of such an intersection in action can be viewed here, . An explanation of how such an intersection can be implemented using V2X can be found in the link I provided above namely, . BTW, when I posted the first link on an engineering forum, one of the guys commented that we would now move from death by car accident to death by heart attack!

PBEndo | 20 September, 2013

@Andrzej1 I sincerely hope you are right since I think we will all be better off when we move to completely automated transportation. I still think the transition will be rough, though. Look at how long it took for the transition from analog to digital TV. Some people are not happy with change. Of course a Teslamotors forum is not the kind of place to find those people ;)

Hopefully, even before human controlled vehicles are mandated, they will be so much better at driving that they can avoid accidents with the small number of human drivers remaining on the roads.

PBEndo | 20 September, 2013

@Andrzej1 Maybe the cars will be designed with passengers facing backwards so it won't be so scary!
Or, replace the windshield with a large monitor showing youtube viral videos. If the computer is driving, the windshield is is unnecessary.

Andrzej1 | 20 September, 2013


Actually some designers have been talking about that in an effort to encourage better socialization within the vehicle.

No doubt steps can be taken even now to reduce accident rates caused by human-driven vehicles. At present there is a large-scale V2X trial going on in Michigan. If the results are positive, one can expect DoT to issue a V2V mandate next year for all new cars starting in 2018. Advocates are claiming a 75% accident rate reduction for human driven vehicles if this mandate is adopted. Let's hope it is and lets start saving thousands of lives yearly. It really is a disgrace that DoT has sat on the spectrum allotted for this purpose for over ten years!

Brian H | 20 September, 2013

Side streets and less-travelled highways will be last to get automated, and will hence require/permit "non-automous" driving much longer -- many years, likely.

How dangerous will drivers be who are competent only to "drive" autonomous cars?

Nagle | 24 September, 2013

Between manually driven cars and fully autonomous cars lies a "deadly valley". The automatic system is able to drive the car by itself - some of the time. But the automatic system isn't able to handle some unexpected hazards. So the driver must be constantly alert and able to take over instantly. That's the Deadly Valley.

Audi's upcoming "Traffic Jam Assistant" works like that. It's able to drive bumper to bumper in a traffic jam. When the situation changes, the automation gives up. A loud beep and a warning message tell the driver to take over.

So the driver has to pay attention to the road while not driving. That is not going to work. It's also somewhat pointless. Why pay for an autopilot if it can't do the job?

Worse, an almost-automatic driving system is potentially hazardous if used in a city with pedestrians and bicycles present. The Audi driver isn't supposed to use it that way. But somebody will. That's not good enough.

On top of that, the autopilot must be safely operable by people who are drunk, drugged, angry, or not too bright.

That's why a 90% solution isn't good enough. When the system reaches a situation it can't handle, it has to be able to get the vehicle stopped at the side of the road by itself. The driver may take over before that happens if they choose. But safety cannot depend on driver alertness during automatic operation.

(I've done some work on automatic driving; I headed a DARPA Grand Challenge team. So I have some idea of what's involved.)

Brian H | 24 September, 2013

Yeah, 90% is hard to imagine. 50% is easier.