Will Tesla Motors Introduce A Massive Autonomous Car Sharing Program?

AUG 21 2015 BY MARK KANE 20

Tesla Prototype Charger

Tesla Prototype Charger

When Uber CEO Travis Kalanick stated that his company would buy 500,000 Teslas if they were autonomous, most of the public focused on how many cars Tesla could sell to Uber.

It’s obvious that sales of 500,000 Model III to a company like Uber is completely out of the question if you can develop you own system of any kind (taxi or car sharing or a combination of both) and expand the business to a much higher level.

Signs of this are coming almost daily as there are more and more autonomous features implemented in Autopilot and the recently presented prototype of a robotic charger.

If you take a look at Apple, who also is working on electric and autonomous cars, the name of the project Titan isn’t accidental, because stake of the future is autonomous cars, workhorses that drive 24/7.

It’s possible that Tesla will try to introduce an autonomous taxi or car sharing (name it as you want) by end of this decade, with Apple being the main competitor.

Thing is hanging in the air and Elon Musk refused even to talk about this, avoiding the most important question:

“Yet another way Tesla could extend its reach was suggested in Musk’s response to an analyst who asked about Travis Kalanick, head of the Uber ride-hailing business, who had been quoted as saying that if Tesla’s cars achieved autonomy by 2020, Kalanick would want to buy all of them. The question was whether Musk might prefer instead to set up a ride-sharing business of his own.

“That’s an insightful question,” Musk said, and then went silent for a good  while. “I don’t think I should answer it.”

For a company manufacturing electric cars and holding software for autonomous driving, adding taxi features seems easy. You will just call the car, and Tesla will send you whichever model you need and then take you to the destination and on its own connect to charging.

How big a market it could be? Well, it could be colossus, like a nationwide car sharing scheme.

Categories: General

Tags:

Leave a Reply

20 Comments on "Will Tesla Motors Introduce A Massive Autonomous Car Sharing Program?"

newest oldest most voted
Pushmi-Pullyu

At present, Tesla’s driver assist features — hyped as “autopilot”, altho it’s rather far from actually being that — requires the driver be in the driver’s seat, and requires him to occasionally indicate to the car he’s alert and monitoring the drive by touching the controls occasionally. And if my understanding is correct, Tesla’s driver assist features are only fully operational on a divided highway, not on any road where two-way traffic, or stoplights/stop signs exist.

The idea of a self-driving taxi requires fully autonomous cars, as well as State laws permitting operation of motor vehicles without anyone behind the wheel. It also requires cars capable of self-driving not merely on divided highways, but also on ordinary streets, including reliably dealing with stop lights and stop signs.

Will all that happen by 2020? I don’t think it will happen that fast. As I understand it, even Google’s self-driving cars can only operate on some roads, not all of them. And Google is far ahead of any other company or research group in development of self-driving cars.

Mark Hovis

I don’t think autonomous driving can be boxed to a single event or date for both technical and legislative reasons.

Components of autonomous driving have already started to arrive with many manufacturers. To say all aspects of autonomous driving do not arrive before 2020 is pretty safe statement.

For me, the piece I await the most is auto park. Even this subset will come in stages. First stage may only allow the auto to park autonomously once it enters the controlled parking area.

The most important phase for EVs will be the ability to travel from an apartment building to a designated parking/charging are even if it is extremely speed limited. This will bring another segment of the market to the EV fold.

Pushmi-Pullyu

“The most important phase for EVs will be the ability to travel from an apartment building to a designated parking/charging are even if it is extremely speed limited. This will bring another segment of the market to the EV fold.”

Right. Unfortunately, the ability of the car to drive safely on public roads including surface streets in a high-density urban area, requires a very high level of dependable autonomy. Indeed, if the car is capable of doing this, what is left to accomplish?

So it seems to me the ability of the car to self-drive between letting the passenger (no longer a “driver”) out at the apartment building, and a parking lot where it can recharge, will be one of the last things to be reliably accomplished, and permitted by State law/regulation.

This is another case where I’d simply love to be proven wrong. As you note, M Hovis, this ability in self-driving cars is really needed to help promote PEV adoption.

Heisenberght

I agree that it’s close to impossible to predict a certain date for when fully autonomous drive will arrive. One has to divide strongly between:

technical feasability:
Take into account that processors, memory, sensors etc. tend to become better in a really fast pace:
Look back to 2010 and what was possible with processors, ssd, memory and sensors in that year.

Sensors: As autonoumous driving is seen as a huge market, more and more suppliers put enormous effort in development. They get cheaper and better and more integrated really fast.

Maps: Also a lot going on here. With increasing amount of sensor-equipped cars, mapping becomes faster.

Software: Look at the job-market, there are quite some job offerings looking for people to develop better algortithms (pedestrian detection seems to be quite hip at the moment…). That should push development faster than in the last years. Here I see the biggest uncertainty in predicting when tech. feas. will be achieved. Just one REALLY GOOD algorithm could change things quickly. And such things happened before. As more and more programmers are working on that subject, the chance that such a breaktrough appears is increasing.

Legislation:
I’ll leave that discussion to the others 😉 Not too much logic in there…

Pushmi-Pullyu
Thanks for the thoughtful analysis, Heisenberght. I don’t think we need any improvement in computer hardware; I think what we already have is quite sufficient to control autonomous cars. Sensors? Well, some development may be needed in terms of placement, or in optimizing the operation for use in this application. I think the basic tech is already there… altho as Elon Musk noted, there are edge cases where the sensors have a hard time “seeing” the edge of the lane near dawn or dusk, in certain conditions. So perhaps some improvement is needed in driving in the rain or other bad weather conditions. Software? Yeah, that’s where the real development is needed. Safely operating a motor vehicle in all conditions is an impossible task. The realistic goal is to produce a self-driving car which has a lower accident rate than the average human-driven car. The basic problem, in terms of computer programming, isn’t all that difficult. It’s merely a matter of collision avoidance, lane-keeping, staying within the speed limit, and noting the presence of stop lights and stop signs. Compared to the operations and math required for some of the more sophisticated computer games on the market, all that isn’t… Read more »
Pushmi-Pullyu

Here is Musk’s actual tweet:

“Final corner case is dealing with low contrast lane markings (faded white on grey concrete) while driving into the sun at dusk”

— July 31, 2015

Heisenberght

Add different wavelenght to the sensor.
Change angle of the sensor.
Add sunlight filter to the sensor.
Add adaptive sunlight filter (LCD) to the sensor.
Light up the difficult spot on the street with laser light (if you are allowed to).
Use additional data from another sensor in these cases. Use predictive algorithm to guesstimate if there is a lane marker.
Add lane marker to global map data at those times when your sensor can see it.
Don’t drive in the evening.
Drive another way in the evening (most drivers rely more on their navigation system than on their own navigation capabilities, so they won’t notice you are taking a slightly longer way)
I guess they will find a way to solve this…

Heisenberght
I agree with you on the hardware-side, up to the point where it comes to real-time-object-classification as you said: “Detours, potholes, objects in the roadway, pedestrians, bicycle riders, a child’s ball bouncing into the road, another car…” are all objects which have to be classified in real time and let me add animals, waste, leafs – not LEAFs 😉 . Up to now this seems to be a bottleneck. I also agree on the point that basic driving features are no problem on the software side. The goal has to be: Merge the datastreams from different sensors (LIDAR + RGBD + “whatever there may come”) There is some trend upcoming that the sensor itself has the object classifaction algorithm integrated in it’s electronics, so the data streams would then already be classified objects. Having different data streams would provide additional security, the main program can for example decide as follows: it’s quite dark already, so I better listen to the LIDAR, because RGBD might tell me something wrong. Error rate will drop by huge means if you merge different sensor data streams. (By the way, does any of the autonomous concept cars include a sound-sensor (say: microphone), that would possibly… Read more »
David Schurig

Think about drones!

I agree that the edge cases won’t be robustly solved autonomously for quite a few years. But…how many vehicles could a remote driver handle if she only had to deal with sporadic problems from the edge cases? 5, 10 20?

Heisenberght

That’s funny, I just thought about drones… Please don’t read my mind OK? 😉

Well a drone that follows the car would need at least ludicrous-mode, so I prefer a set of drones creating a grid above all roads (might be expensive and you need a lot of superchargers for those drones to keep them in the air 24/7)

so I changed my mind on smart infrastructure:

add box with a cheap set of cameras to each bridge, running a low-cost low energy (RASPI or something) using solar modules (something in the 50W range buffered with 50Ah LI-Ion) thing just to determine how many cars are travelling at what speed and send that data via UMTS or something to a server to provide real time traffic data to everyone. Such a unit would cost something like 250$ or maybe 300$.

Could drastically improve flow of cars and also effectively increase the cars line of sight, as an accident or traffic jam would be reported imediately.

Pushmi-Pullyu

One thing about autonomous vehicles which I think has not been sufficiently discussed, or realized by the public, is the benefit of interconnectivity. Self-driving cars are not going to operate independently; they’ll wirelessly share data with each other, and probably with centralized traffic computers. Traffic jams will be instantly reported to the network, and traffic immediately diverted to alternate routes. When cars in close proximity can communicate with each other, they can signal their intended vectors to other cars, and thus virtually eliminate any chance of collision, except in cases of malfunction.

If and when only autonomous cars are allowed on the roads, and no human driven cars are allowed, then we won’t even need stop lights or stop signs. Cars will simply “agree” on right-of-way using some predetermined algorithm.

jkw
The difficult edge case isn’t dealing with other cars. Cars are big, easy to detect, and have a limited number of ways they can move. I’m pretty sure that current autonomous driving can already do way better than most people at not hitting another car, even when the other car is being driven recklessly. The new safety feature that is likely to become standard is the automated braking assist systems that do avoid hitting other cars pretty reliably. The difficult edge cases are things like knowing when something small needs to be avoided and when it doesn’t. A plastic bag blowing in the wind can generally be ignored, but a toy that could look almost identical to the sensors is likely to be followed by a child. There are also times when people do strange things that won’t be anticipated in the programming. People can tell when the stopped car in the right lane is planning on turning left as soon as the light turns green, but it would be very hard to make a computer understand when that is likely. People can see a barely visible person peaking out around a large car and know that the person is… Read more »
Heisenberght
“People can see a barely visible person peaking out around a large car and know that the person is likely to step out into the road, while the sensors on a car probably won’t even notice the person until they are in the road.” Well that should for example depend on the location of the camera. Imagine a camera mounted on top of the car, from that angle the camera could even see the person, when it is still covered for the driver. Imagine two cameras on the upper edges of the windscreen. Imagine additional cameras in the front of the car. Don’t forget, cameras are cheap as hell! So once you have that data (before the driver has it) you just need an algorithm to detect the person (classify it as a human beeing) Face detection algos could help here. Adding infrared sensors could increase reliability. That algorithm will eventually be even faster in realizing that there is a human beeing than a possibly distracted driver will. So I think that this won’t be a problem for long… The toy and the bouncing ball are definitely harder to solve. But once you have been at a busy road with… Read more »
Pushmi-Pullyu
Hiesenberght said: “…you just need an algorithm to detect the person (classify it as a human beeing) Face detection algos could help here.” I’m not sure what purpose would be served in optimizing the software to recognize a “person”. It’s a sizable moving object, so the car should avoid colliding with it. That would be true if it’s a human, a large dog or small deer, or a box blowing in the wind. Recognizing it as a human being would be a very hard task, and impossible to perform reliably. What if the person is looking away, never presenting a face to the camera? What if the person is wearing a costume or voluminous clothing, so not recognizable as a human shape? What if it’s someone in a wheelchair, or a child riding a tricycle, so the basic profile is altered? Now, I can see the argument that in theory, it would be better to assign a higher priority to not hitting a person; the priority would be something like “avoid hitting this object designated ‘human’ even if risky emergency maneuvers are necessary; even if it’s necessary to crash into a parked car to avoid collision”. But that’s the sort… Read more »
Kosh

So what would they do when at 5 pm everybody in SF calls for a car ride home? They certainly would NOT keep peak n cars around idle the rest of the day…

force people to carpool on the autonomous system?

Heisenberght

They could increase price per mile for the drive at peak hours.

Someone out there

Tesla is in a great position to do so with their own cars, their own charging network and effectively their own power generation (SolarCity).
The real question however is whether fully autonomous drive will be legal by 2020. I doubt it.

Driverguy01

He talked about sending hundreds of mini satellites with spaceX, he could use those for real time guidance instead of whats available today…

Delta

‘ P100, Engage ludicrous mode please’

‘I’m sorry, i can’t allow that, Dave.’

michael

I think that self driving cars aren’t going to happen in the way most people assume. The popular assumption is that they’ll take over the autonomous, independent means of transportation we currently have, in a comprehensive way, all at once.

What I predict will happen is that they will take over low to medium speed taxi service in city centers like London, Tokyo and New York. Then they’ll expand to faster point to point taxi services back and forth to specific airports, outlying event centers and attractions. Only after they’ve been proven in those roles will the systems be activated fully in personally owned cars.

This staged introduction is the only method that I can currently envision that details a gradual introduction to build confidence in their abilities while still remaining in profitable service at every step of the way, which is why I feel certain that this is the way we will transition to autonomous vehicles.