Elon Musk: Tesla Autopilot 1.01 Coming Soon

OCT 24 2015 BY ERIC LOVEDAY 28

Elon Musk Tweets

Elon Musk Tweets

Tesla Model S w/ Autopilot

Tesla Model S w/ Autopilot

According to Tesla CEO Elon Musk, a new, improved version (1.01) of Autopilot for the Model S is already in the works.

Called Autopilot 1.01, Musk says the improvements include:

“…curve speed adaption, controller smoothness, better lane holding on poor roads, improved fleet learning!”

Musk notes that this updated Autopilot is “coming soon,” but gives no exact timeframe for its rollout.

Musk was asked via Twitter if a retrofit would be possible for early Model S electric cars. His answer stops short of being a flat out no, but does suggest it just simply wouldn’t be worth the price due to the amount of work required.

Categories: Tesla

Tags: , , ,

Leave a Reply

28 Comments on "Elon Musk: Tesla Autopilot 1.01 Coming Soon"

newest oldest most voted

Soon they are going to leave others far behind in auto pilot game. This fleet learning us going to give them as much of an edge as Google got from search. A positive cycle where fleet learning fine tunes auto capabilities to such an extent that more and more people buy it’s cars leading to even better capabilities.

If I understand correctly, Tesla has already left others far behind as far as what has actually been deployed for use by drivers not specifically trained for its use, as is true of Google’s self-driving car project. From online discussion, I think that only Mercedes has deployed a lane-keeping feature in one or more of its luxury car models, and there seems to be a pretty firm consensus that Tesla’s Autosteer works a lot better. Now, that’s not to say that Tesla’s rather optimistically named “Autopilot” is as advanced as Google’s system for driverless cars, which as I understand has advanced to the point that newer prototypes don’t even have a steering wheel! But Google hasn’t taken the bold step of deploying its sensors, software and hardware into cars driven by Joe Average. I personally think Autosteer (Beta) shouldn’t have been released “into the wild” without further improvement, but at least so far we haven’t seen any reports of accidents caused by the system — which, statistically, seems to be evidence that it’s already at the point at which using that is safer than letting a human steer. Yeah, there have been a few reports of near-accidents using Autosteer (Beta),… Read more »

They needed the test data from Autopilot 1.00 to test against.
From people driving all over the world, in all kinds of roads, to the maniacs letting the autopilot drive a Model S @ 90mph through some crazy turns.
Its been humored that whenever the autopilot encounters interesting situations the car beams sensor data back to the factory, which gives developers data to back test their improvements against.
Bur certainly ballsy. Typical Musk.

AutoPilot (AutoSteer) should not be blamed for any accidents as the driver is responsible and is supposed to be keeping their hands [lightly] on the steering wheel. Being hit by another car while the Tesla is in a blind spot of car seems like an obvious time when the handfree driver will not have time to react and will be in an or cause an accident (move to other lane as a reaction)

scottf200 “AutoPilot (AutoSteer) should not be blamed for any accidents as the driver is responsible and is supposed to be keeping their hands [lightly] on the steering wheel.” I absolutely agree; adults are supposed to be responsible for their own actions. But that’s not going to stop the media from going into a feeding frenzy the first time someone gets killed or maimed in a car being steered by self-driving software, no matter how many disclaimers the auto maker used to warn the driver (or non-driver, as the case may be), and no matter if the driver was ignoring clear warnings. Disclaimers are not going to stop Tesla’s reputation from taking a huge hit, if and when there’s a fatal or crippling accident involving a car steered by Tesla Autosteer, and with our litigious society, it seems all too likely that a jury would feel so much sympathy for a victim that they would award him (or them) millions of dollars of Tesla’s money. There is the ideal, and there is the reality of the world we live in. Ideally, all drivers of Model S’s would heed the warnings about Autosteer (Beta), and would engage it only on freeways, and… Read more »

Please tell me what these “updates” consist of, and what information would improve the autopilot.

Evolution is a beautiful thing… 🙂

I don’t think what Tesla has done with the Autopilot software upgrade is as groundbreaking as some of the posts here indicate. Both of my Fords provide user habits to Ford through Sync-My-Ride. My 2015 Ford Fusion Energi has Adaptive Cruise Control, Lane Departure Warning and Parking Assist. My Fusion should have all the sensors and controls to achieve autopilot performance with nothing more than a software change. If Ford doesn’t do it, I think it’s only a matter of time before an aftermarket company provides a software upgrade to autopilot features. Many cars have the same autonomous features that my Fusion has. What I think that Tesla did that is huge is that they got lane centering past regulators. I didn’t think lane centering would get past regulators for another ten years if ever. Now the proverbial Pabora’s Box has been open. For better or worse we are on our way towards fully autonomous vehicles. I think we are going to see a lot autonomous features coming out even on low end models in the very near future. One of the reasons regulators are probably allowing autonomous features is that they perceive the features as making the cars safer… Read more »

Does your Ford steer itself?
The software is the KEY. Noone made one yet. But some have started and only one have realeased it! Tesla.

Yes my Fusion steers itself both in lane departure warning and in parking assit.

Lane departure detection is nowhere near as sophisticated as autosteer. If you used that technology to steer, you’d be bouncing left and right between the lines, and the car would keep reacting way too late. It’d look like a drunk was driving.

Autopilot needs to know where the lanes are well ahead of the car, not just when the car gets near them.

*lol*…

Plenty of manufacturers have “made one yet” and realeased it.

Basically every manufacturer has models with autosteering.

But even now that Tesla deployed late in the game that is not what matters. What matters is how good you make it and it looks like Tesla’s is one of the best.

Considering how easily they can update it and how much they now are learning from tens of thousands cars on the roads it is safe to say that they soon will be the best one actually on the road in the hands of normal owners and with a big margin.

So since you’ve had you’re Fusion, how many software upgrades have you been offered? And have you bothered to take the car in to get them?

Since 2012 Tesla has had 7 major releases, with many more in between. And at no inconvienience to the owners. This pace is even aggressive for companies like Apple. Tesla innovations go far beyond anything in the market today. And the pace and strategy they are taking will only accelerate their lead in so many ways. To be passing all competition in so few years, and with adaptive network learning capabilities, Tesla will soon even catch what Google has achieved with self driving. Just in time to share it with the masses via the Model 3…

+1

My Fusion is a 2015 model and I haven’t had any opportunity to try to get the software upgraded. Ford isn’t very proactive when it comes to software changes, that’s why I said an aftermarket company might have to do the software changes to give me lane centering. I did take my 2013 FFE in to the dealer recently and ask them to update all the software they could. They updated the software in a couple of modules but the only improvement I noticed was that my voice activation works a little better.

+1 I took my Jaguar to the dealer for an update patch because the CD player failure and It took 1 day, no way they will keep updating self driving software that way. Go Tesla

Mikael
Please tell me what information will the car send back to Tesla that is so important or contains more than can be found using Google Earth Photos.

Frank, it is the exact position of the car on the map. The problem with GPS is that it is not accurate enough for lane keeping. Actually accuracy of GPS is quite horrible, about 10 meters. Therefore GPS alone is more or less useless for lane keeping.

Therefore Tesla must use cameras for lane keeping, but markings are not always clear enough. With fleet learning, Tesla can identify those sites that need more attention for software to figure out the exact position of the car.

Jouni, do you mind if I some additional thoughts to bear on the GPS discussion? Since GPS was really invented for use by Military, and has been borrowed for use by Civilian Applications, do you remember that common civilian GPS is a downgraded signal? And – since Aircraft have started using this civilian quality signal, and wanted better precision for instrument approaches in bad weather, they have implanted an aerospace ‘upgrade’ called ‘Differential GPS’ that greatly improves the accuracy if you have the correct ‘DGPS’ receivers installed! Differential GPS uses a extra GPS Receiver, that I located on a highly Precise Ground Location and sends out ‘corrected’ signals to DGPS Receivers to give them this improved Precision. For Aerospace, these are typically on, or near major Airports! So, if Tesla wanted to take advantage of these signals – they could move to better DGPS Receivers (if they have not already!), and they could also install some additional Points to cover additional areas of interest, should the investment warrant it, but – they could share the investment costs with the National Aerospace development too, as it would also add benefits for them! Per Wikipedia – “Differential Global Positioning System (DGPS) is… Read more »

The reason civilian Gps is poor resolution is so that the system cannot be used to guide a homebrew missile.

Yes, but since Differential GPS was invented to improve navigation in important places like major seaports, why can’t it be used in a homebrew missile too?

I don’t know enough about GPS to have a well-informed opinion, but a friend of mine says that he has a GPS device which takes the signal from ten different GPS satellites and averages the data to deliver an order of magnitude improvement over standard civilian GPS. As someone already noted above, the military grade is rather better.

But cruise missiles use terrain following radar and a digital map to get the within-two-feet precision they need for “smart” targeting. I think Jouni Valkonen is correct; I think GPS simply isn’t precise enough for lane-keeping. And that goes right along with Google using vans to scan the roads where their self-driving cars are intended to drive.

I’d be interested to know just what kind of data, and what level of detail, Google uses — and I would guess that
Tesla is beginning to use — for mapping roads with sufficient precision to ensure no self-driving car ever wanders out of its lane. Looks like Google has achieved this, and apparently Telsa is working to achieve the same thing. But even Google probably hasn’t yet scanned every public road in the USA, let alone the entire world.

PP – “But cruise missiles use terrain following radar and a digital map to get the within-two-feet precision” – and Tesla uses forward looking Camera’s and Radar, Plus I expect on the average Freeway in the World where Tesla’s drive, there is more than one Tesla now, so the ‘Learning’ part gains an edge over time. So – if you combined that with DGPS, or aviation’s ‘WAAS’ variant, you could increase the GPS part of the data to so much closer than the 10 – 15 meter typical precision. Even if it was 20 cm instead of 10, when combined with the other two sensors, could do much better.

Even if yours was the only Autopilot equipped Tesla on a given road, if it was constantly monitoring, it would gradually over time and trips down a given road, get better data to work with, I think – and that seems to be a part of the ‘Learning’ aspect. (Sorry, not stating a fact, just my understanding of the idea!)

The “downgraded signal” thing is (was!) called “selective availability” and has been turned off since 2000. http://www.gps.gov/systems/gps/modernization/sa/

Is anyone reporting that autopilot is showing that it is learning? For example, there’s at least one video showing a TESLA trying to exit the freeway on an exit when the driver had no intention of exiting. Does the car actually learn not to take that exit after driving that route a few times?

Yes, there are multiple reports on TMC of learning apparently happening, and specifically with spurious exiting.

It’s not clear (to me, anyway) whether learning is driver-specific or overall fleet learning.

Even 1.0 introduced some significant degree of curve speed adaptation to TACC (even when autosteer isn’t engaged), a nice surprise I may add, since Tesla didn’t even mention it. I wonder what 1.01 will add in that regard?