Is the hardware of Tesla sufficient for Level 5 autonomous driving?

Is the hardware of Tesla sufficient for Level 5 autonomous driving?-autonomous

As a great technical, electromobility and Tesla fan, I have been following Elon Musk for more than 10 years and devouring any news and YouTube videos to the Tesla and its innumerable fans. But just the statements for autonomous driving made me really a bit suspicious. How should a camera system with so many dead angles, ever can drive self-employed.

Due to the small number of cameras or the non-optimal position of the existing cameras, it is actually “physically impossible” to cover the entire necessary viewing area – without in some situations – to get problematic vigor gaps. How Elon would say: “It’s Just Physics”. To check this and due to my animation experience, I decided to check this in a simple simulation. Thus, in Aftereffects with the Cinema4D renderer, as precisely visualization on the basis of the data mentioned by Tesla, was rebuilt.

A simplified version of a Tesla model 3 and visualized and simulated with cameras (visualized as colored light cone) with the same angles and positions like Tesla on its homepage – but only for the left side of the vehicle to the simulation clear and the complexity keep in the frame. The green light cone simulates the front camera with the largest wide angle and the side camera is visualized by the red light cone.

The result can be viewed in this video on YouTube:

Here it can be clearly seen that by Tesla’s camera placement – especially to the sides – arise in many situations dead angles. So that Tesla gets the necessary view, the model 3 would have to drive very far into the street center and thus hinder the crossing traffic. Exactly this behavior can be observed again and again in FSD-beta (FSD = Full Self Driving) videos of other Youtuber. In the simulation, it can also be seen that the necessary reaction and brake path for the crossing traffic in this situation would be far too low – so in my opinion results in an increased risk of collision.

That’s why I simulated the same situation again and tried to reduce the risk in which I have a further laterally aligned cameras approximately at the height of the front headlight (pink light cone). By this, at least in relation to the laterally crossing traffic, the dead angles are extremely reduced and make a wide retraction in the road unnecessary – whereby the collision risk should decrease.

In addition, this would increase the redundancy – which represents another big problem (especially for admission in Europe).

Meanwhile, more FSD-Beta Youtuber has become aware of the problem that the car has been extinguished in the beta for a long time. The Youtuber “Chuck Cook” has after many problems with the FSD-beta (do not want to turn left) to be attached to front headlights further cameras and shows how to reduce these dead angles and that clean rolls in the street. He has thus tried the same solution proposal as in the simulation here in “real-life”. To see in this video:

In the research I came across more problems that were not treated in the simulation:

Thus, the pages of cameras seemingly only one (relevant for the artificial intelligence) view of 80 meters, which is far too low that crossing traffic (for example, 100 km / h) incl. Reaction time can come to a halt. Even the low resolution of some cameras of only 720p and the poor image quality in the peripheral areas of wide-angled cameras, could lead to view (-fares) and detection issues. The above-mentioned redundancy is of course an additional problem that could become the risk of defective or simplicating pollution of a camera.

In my opinion, the currently existing hardware is not enough to drive real level 5 autonomous and thus enable the FSD taxi service, for example. But Tesla will probably be able to admit, with a view to the stock value and the still something unstable market position, for the time being. But as long as you continue to research and develop a few years, you will not have to do that – but sometime in the next few years, Tesla has to bite in the acidic apple:

If you look at the countless FullSselfdriving beta videos on YouTube, the current hardware in my opinion has already arrived at your limit that one would currently be described as Level 3 partial autonomous driving. Who says: “Tesla has long driven on highways / highways almost fully autonomous”, I say: “Then put a not moving object in the middle of the highway and pray.”One can really do Tesla, the” with-swim “in traffic on the highway. Collisions seems to be given above all with standing / solid objects, thus approaching rapidly – there seems that the current visual detection of dangers has not always sufficient – as one can refer to some accident on the internet.

The former Israeli developer (and subsidiary of Intel) of autopilot hardware 1.0 named “Mobileye”, a system has designed a system with a little more cameras (and an interesting, little data needed mapping system), which in many cities already properly and even in Jerusalem already well adapted (/ pure-urging) in the Care can ride. Here is a 40-minute and uncut drive through Jerusalem on YouTube:

Finally, it remains exciting in this area, Tesla but in my opinion, needs at least one more comprehensive hardware update to achieve your FSD plans. Since Tesla but nowhere of “Level 5 autonomous driving” speaks and also less heard from the FSD taxi service in recent years, a rogue might suspect, the Tesla simply says: “Such the FSD software is now done.”And then simply a system published that somewhere is between Level3 and Level4 – quite an option, if certainly not the most popular.

But now, however, I now see Mobileye wide, even since it is the only company that (as well as Tesla) has collected and analyzed the data millions of millions of vehicles throughout the world for years. Yes, not only Tesla came to the idea and Mobileye as the market-leading manufacturer of current assistance systems, has just installed millions of systems in the cars of many large manufacturers, which have been collecting data for years and traveling worldwide on the roads.

As I said: The autonomous future will come, but it will stay exciting.

About the author: Written by Jonas S.

Related articles

Please follow and like us:

9 thoughts on “Is the hardware of Tesla sufficient for Level 5 autonomous driving?”

  1. Maybe the author should consider that Tesla engineers work above his salary class.
    If the alleged problem with 1-2 other cameras and a better resolution is to be resolved, then he should apply to Tesla, and all “just show how it works”
    Maybe it should be thought about what industries Elon has shown so far what just goes
    The auto industry smiled the idea of an electric car for a long time:

    • impossible
    • no reach
    • Too slow shop
    • No power (can only accelerate 1x, then power is gone)

    What are the today to a model S Plaid?

    The space industry smiled for Space-X Lange:

    • impossible
    • Cheaper does not work
    • Reusable missiles are nonsense
    • Floating Landing Platform does not work

    Who today has the order for the next moon landing?

    The telecommunications companies have a Star Link regularly smiled:

    • impossible
    • way too expensive
    • Only for rich (like satellite phone)
    Who sells today the complete system for $ 499 at a monthly price of $ 99
    Incidentally, Elon Musk has restored with Starlink in various catastrophe areas in the world so the contained combined communication – yes, even recently in Ahrweiler!

    I am also Teslafan and I have my doubt with my doubt, but I would not consider my very limited understanding with regard to autonomous driving as well as that the solution is simply 2 more cameras and a higher image resolution is.
    I’m now waiting for a relaxed, what Tesla will imagine at the Ai-Day and which system makes the race in the future and I believe in the amount of tests growing with any Tesla rather at Tesla than MobyEye!
    Time wants Tell &# 128521;

    Reply
  2. The setting “only what 100% works is allowed to go to the street” annoy me. Then there would be no horses, let alone can give cars. As a driver of a car I do not sit at the front license plate. I sit on the driver’s seat and see something like Teslas camera. And I look at every time in contrast to Tesla only in one direction and react additionally slower. Dear 720 P as a motorist who have a wrong set glasses.

    And eternal focus on Tesla and Mobileye also annoys. These are just two providers. Whether they will be the first to bring the autonomous driving is uncertain. In China, countless Robotaxis drive. It seems to me as if many editors looked only to Germany and the USA. That may probably be because you can not get Chinese news and read. Currently working about a dozen companies on the autonomous driving. Nobody knows their technology and the actual requirements of autonomous driving so careful that he can estimate what these approaches are promising and which are running in front of a previously unparalley wall. If they already knew about this wall, they would have changed their concept and now would not know anything about the wall, which comes to them.

    Reply
  3. I would not accuse a car group that he has not built the entire hardware for autonomous driving, as long as the most important states have not set the conditions for such autonomous driving not even precisely. In extreme cases, the HW equipment is in principle, but it is not enough for the concrete requirements of important states. Tesla is perhaps a good single player, but autonomous driving is just no single sport, but it requires a collaboration of car manufacturers, standardization committees, traffic control system manufacturers and state.
    Problematic is for me, on the other hand, when a manufacturer promises stuff, which you can not even promise. Since the communication of Tesla is very clever. Wrong, or. Exaggerated expectations of naive consumers are promoted, although they do not say that very concrete. Exactly that annoys me since the beginning at this company.
    But you can also see it differently and think who does not think precisely and continuously exaggerates expectations is to blame itself. Tesla just has a lot of uncritical followers because they master the art how to make such groups. Similarly, like Apple for a long time or even successful churches and sects.

    Reply
  4. Numerous videos and reports from the partly personally acquaintances FSD Beta users show that the number of cameras at Tesla for autonomous driving on SAE4 are completely sufficient. Focusing on sensors has already shown on the subject of Lidar and Radar that a layful understanding (no attack) to the wrong conclusion leads the sensors a problem or topic are there the algorithms nd the conflicts arise due to too much different sensors. I have made an interview with Steven Peeters who evt. be interesting for one or the other and another with a German channel follows in August. Furthermore, there are various articles from me on the topic.

    On the topic of standing objects that are not noticed by a Tesla vehicles is to be noted that this statement is faulty. Tesla gives to his system as well as all the other other in the market systems currently no standing objects on highways or highways dodged. The other manufacturers of autonomous systems make this as well! The background is that you do not have full braking because of an unclear object on a highway as long as the system is still regulated as SAE2 (driver with the responsibility). Here is the responsibility to the driver and that is not a mistake of the system, but as ‘Instruction Manual’ so defined. Once Tesla is going to L4, this circumstance will change.

    I could write more about the article because it is full of false conclusions and assumptions but would like to leave this commentary at this point.

    Unlike such an article should not appear on electric car news but be rejected by the editor. This is not about the criticism of the AP or FSD, which the author of almost all places either confused or not differentiated (consciously or ignorance?), but poor bad research, lack of know-how and disseminate false information on the topic in general.

    Reply
  5. Before being discussed here for a long time over autonomous driving, Tesla should first improve basics, such as sign recognition, for example.

    Last week I went outside of closed villages over a bridge. There was no signposted speedimit, but a limitation of the vehicle mass to 30 tons. The car has ignored the “T” behind the “30” and the autopilot has thrown the anchor. What should the Tesla?

    At the beginning of this week I was on the Inntalautobahn in Austria. That IG-L limits do not apply to clean electric vehicles, apparently did not speak to California. Similarly, the autopilot also reacts to temperpolimits in wetness, though radiant sunshine prevails.

    I know from other manufacturers (Audi, Mercedes-Benz) that in which the latest systems for sign recognition work better.

    Reply
  6. As an author I would like to complement one more here:

    The simulation is an experiment based on the physical laws based on the data provided by Tesla … no more and no less.

    I’m a great Tesla fan, can also distinguish between AP and FSD (also knows that on the highway to the AP is changed in the FSD beta), contact anywhere of Lidar sensors or radar and does not want to hinder the development, nor Tesla misappropriate.

    I just know here on a physically existing problem and give my suggestion and put my opinion on solving the problem.

    These screeching here, scolding and decaying chicken stall decayed in undertaking, please look at yourself in the mirror.

    I just turn to a partial problem, no more and no less – but what’s going on on reactions, actually occupies only the picture that most of us have Tesla fanboys … and yes, I belong to it – but I also recognize mistakes And that is one of a development and improvement. Otherwise nothing would be better … So please love Tesla fans: get you again.

    Reply
  7. The question is: how does a person master these situations? I suspect that people who are always standing on such confusing point become more and more nervous the longer they stand there and sometime easy to pull out. That’s good, or not. For example, if a vehicle comes from the opposite direction, but what you have not heard because you only focused on one page. This can not happen to a vehicle with all purpose. And a Tesla with its strong acceleration could have just had to pull out in the situation, instead of standing again.

    The suggested camera would certainly defuse this problem. But whether she is really necessary for Level 4/5, will show itself. If necessary, the dare could also decide to turn on the right and to address if the situation is unless otherwise mastered. We speak of a fully autonomous vehicle. And that would also be a way to master inadequacies of the system.

    I am neither Tesla fanboy nor would someone who buys a vehicle with FSD. If it would be so far I would only call such vehicles via app

    Farnsworth

    Reply
  8. I also see MobileEye front … Tesla Hechelt (risky and ready, see beta program) after ..

    Me wants to start a robotaxi service in Munich in the middle of next year? What has Tesla to offer?

    Reply

Leave a Comment