'Apple Car' needs Machine Learning to make driving decisions fast enough

Jump to First Reply
Posted:
in Future Apple Hardware
Apple is planning to use Machine Learning in the "Apple Car," specifically because current processors are not fast enough to autonomously make key driving decisions without the technology.




It was already expected that Apple would use Machine Learning (ML) in the expecteed "Apple Car," and not least since the company's AI chief John Giannandrea was placed in charge. Now, however, a newly-revealed patent explains just how ML will be used - and also why it is needed.

"Evaluating varying-sized action spaces using reinforcement learning," is concerned with the car benefiting from ML The idea of a car learning from its own mistakes is a little scary, but this is more like the car getting to use accumulated data from all such cars.

It's all to do with how decisions taken at the wheel may have to be extremely fast. Even a correct decision about, say, a lane change or avoiding a collision, could be fatal if it is not completed quickly enough.

"Until relatively recently," says the patent, "due to the limitations of the available hardware and software, the maximum speed at which computations for analyzing relevant aspects of the vehicle's external environment could be performed was insufficient to enable non-trivial navigation decisions to be made without human guidance."

The phrase "until relatively recently," seems to imply that hardware and software are getting better. They are, but Apple then says it's still simply not enough.

"Even with today's fast processors, large memories, and advanced algorithms," it continues, "however, the task of making timely and reasonable decisions... of the vehicle's environment remains a significant challenge."

The patent talks about the complexity of autonomous decision making which is "based neither on excessively pessimistic assumptions, nor on excessively optimistic assumptions." Then cars might be able to drive themselves, but they'll never drive alone - so the "unpredictable behaviors" of other drivers in other cars are a factor.

Plus the real world is a lot messier than any test environment, so Apple also notes that autonomous driving decisions will have to be made even when there is "incomplete or noisy data."

Over 17,000 words, the patent describes situations to do with the car's "action space." That's the time and distance within which the car has to make its decisions.

"In some states, such as when the vehicle is traveling on a largely-empty straight highway with no turns possible for several kilometers or miles," continues the patent, "the number of actions to be evaluated may be relatively small; in other states, as when the vehicle approaches a crowded intersection, the number of actions may be much larger."

Detail from the patent showing one example of decision making
Detail from the patent showing one example of decision making


In each case, the car's systems have to determine "the current state of the environment" around the vehicle. Then it may need to identify "a corresponding set of feasible or proposed actions which can be undertaken."

An action could be "turn left," or "change lanes." In at least some cases, ML can be used to help the car assign a number or value to each possible decision, and then determine the best course of action.

"[For example,] multiple instances or executions of a reinforcement learning model may be employed at the vehicle to obtain respective value metrics for the actions," says the patent, "and the value metrics may be used to select the action to implement."

This patent is credited to two inventors, Martin Levihn, and Pekka Tapani Raiko.

Levihn's previous related work includes a patent for a "behavior planner" for a vehicle, another autonomous decision-making system.

Read on AppleInsider
alexsaunders790

Comments

  • Reply 1 of 14
    sflagelsflagel Posts: 869member
    When a car is able to successfully navigate the Elephant and Castle Roundabout in London, at night, during rush hour, in December, in the rain, then it shall be deemed safe.
    lkruppjas99command_fbaconstanglolliverpatchythepirateapplguyOctoMonkeyJWSCwatto_cobra
     11Likes 0Dislikes 0Informatives
  • Reply 2 of 14
    Isn't IBM selling Watson?
    watto_cobra
     1Like 0Dislikes 0Informatives
  • Reply 3 of 14
    dk49dk49 Posts: 287member
    There were recent rumours of a "C1" chip developed by Apple to run the car. Seems like it didn't suffice. 
     0Likes 0Dislikes 0Informatives
  • Reply 4 of 14
    lkrupplkrupp Posts: 10,557member
    I am in agreement with other critics of self-driving cars. As long as humans remain in control of cars ,machine learning, artificial intelligence, sensors, etc. will be no match for the eccentricities, poor judgement, inattention,  and road rage of homo sapien  drivers.

    On a more positive note, I read an article about an Israeli company who has contracted to build a one mile stretch of highway in Michigan with charging coils embedded in the pavement to allow wireless charging of EVs while they drive over the road. Of course the EVs will also be equipped with coils as well. If this turns out to be a practical solution to the range issue of EVs it will change everything.
    jas99command_fbaconstangcurrentinterestJWSC
     3Likes 0Dislikes 2Informatives
  • Reply 5 of 14
    dk49dk49 Posts: 287member
    lkrupp said:
    I am in agreement with other critics of self-driving cars. As long as humans remain in control of cars ,machine learning, artificial intelligence, sensors, etc. will be no match for the eccentricities, poor judgement, inattention,  and road rage of homo sapien  drivers.

    On a more positive note, I read an article about an Israeli company who has contracted to build a one mile stretch of highway in Michigan with charging coils embedded in the pavement to allow wireless charging of EVs while they drive over the road. Of course the EVs will also be equipped with coils as well. If this turns out to be a practical solution to the range issue of EVs it will change everything.
    This concept has already been tried before. And it didn't quite scale. 
    StrangeDayswatto_cobraalexsaunders790
     3Likes 0Dislikes 0Informatives
  • Reply 6 of 14
    elijahgelijahg Posts: 2,876member
    lkrupp said:
    I am in agreement with other critics of self-driving cars. As long as humans remain in control of cars ,machine learning, artificial intelligence, sensors, etc. will be no match for the eccentricities, poor judgement, inattention,  and road rage of homo sapien  drivers.

    On a more positive note, I read an article about an Israeli company who has contracted to build a one mile stretch of highway in Michigan with charging coils embedded in the pavement to allow wireless charging of EVs while they drive over the road. Of course the EVs will also be equipped with coils as well. If this turns out to be a practical solution to the range issue of EVs it will change everything.
    There would need to be vast amounts of copper embedded into the road to make this work. It'll never be feasible, not unlike solar roads. 
    patchythepirateOctoMonkey
     2Likes 0Dislikes 0Informatives
  • Reply 7 of 14
    cg27cg27 Posts: 223member
    dk49 said:
    lkrupp said:
    I am in agreement with other critics of self-driving cars. As long as humans remain in control of cars ,machine learning, artificial intelligence, sensors, etc. will be no match for the eccentricities, poor judgement, inattention,  and road rage of homo sapien  drivers.

    On a more positive note, I read an article about an Israeli company who has contracted to build a one mile stretch of highway in Michigan with charging coils embedded in the pavement to allow wireless charging of EVs while they drive over the road. Of course the EVs will also be equipped with coils as well. If this turns out to be a practical solution to the range issue of EVs it will change everything.
    This concept has already been tried before. And it didn't quite scale. 
    And definitely not in pot-hole miserable MIchigan where new roads crumble after only a few years.  Would get extremely expensive to replace very expensive copper coils.

    Not to mention, in Detroit the jackhammers would come out to dig up the copper and sell it.
    edited February 2022
    JWSCwatto_cobra
     1Like 0Dislikes 1Informative
  • Reply 8 of 14
    cg27cg27 Posts: 223member
    elijahg said:
    lkrupp said:
    I am in agreement with other critics of self-driving cars. As long as humans remain in control of cars ,machine learning, artificial intelligence, sensors, etc. will be no match for the eccentricities, poor judgement, inattention,  and road rage of homo sapien  drivers.

    On a more positive note, I read an article about an Israeli company who has contracted to build a one mile stretch of highway in Michigan with charging coils embedded in the pavement to allow wireless charging of EVs while they drive over the road. Of course the EVs will also be equipped with coils as well. If this turns out to be a practical solution to the range issue of EVs it will change everything.
    There would need to be vast amounts of copper embedded into the road to make this work. It'll never be feasible, not unlike solar roads. 
    Yeah, agreed, lots of things seem great and feasible on paper, like Better Place’s swappable batteries tried ten years ago.  Swappable is happening in certain parts of China now, and for ebikes, but still skeptical it’ll work everywhere.
    elijahgwatto_cobra
     2Likes 0Dislikes 0Informatives
  • Reply 9 of 14
    lkrupp said:
    I am in agreement with other critics of self-driving cars. As long as humans remain in control of cars ,machine learning, artificial intelligence, sensors, etc. will be no match for the eccentricities, poor judgement, inattention,  and road rage of homo sapien  drivers.

    In your view, what is the standard that AI has to meet?  Does it have to be perfect? Does it have to be as good as the very best human driver? Does it have to be as good as an average driver?

    I expect current AI tech clears that third target and am confident that it will meet the “best driver” standard within a decade. Perfection is a straw man. 

    As you suggest, human drivers make the challenge much harder, but such are the challenges that we human drivers face today. 
    watto_cobra
     1Like 0Dislikes 0Informatives
  • Reply 10 of 14
    sflagel said:
    When a car is able to successfully navigate the Elephant and Castle Roundabout in London, at night, during rush hour, in December, in the rain, then it shall be deemed safe.
    Why set the bar there? ... we already let humans attempt that... and they're obviously not 100% capable of it.
    I suppose the question becomes: "just how often does it need to do this successfully?... what error rate are we willing to accept?" I would think that as long as it's a single tic better than the human rate, then it should be allowed on the roads. (Else, we should quit letting humans drive as well.)

    watto_cobra
     1Like 0Dislikes 0Informatives
  • Reply 11 of 14
    thttht Posts: 5,801member
    In each case, the car's systems have to determine "the current state of the environment" around the vehicle.
    This is a big crux of the matter. There really needs to be a full suite of sensor data with compute hardware and software to interpret it every 5 seconds or so. It is a dynamic interpretation, so they have to look at it holistically for a few seconds. Object recognition for images or a single frame of video just needs to work for that one frame. You need multiple frames to figure out an objects' speed and direction.

    Don't like that they are waiting for the hardware and software to get good autonomous driving to ship an EV. If they built a "dumb" EV, they can get all the supply chain and manufacturing work out, especially their battery supply chain and sales/service channels. They can increase their driving data by orders if magnitude.
    watto_cobra
     1Like 0Dislikes 0Informatives
  • Reply 12 of 14
    sflagel said:
    When a car is able to successfully navigate the Elephant and Castle Roundabout in London, at night, during rush hour, in December, in the rain, then it shall be deemed safe.
    Why set the bar there? ... we already let humans attempt that... and they're obviously not 100% capable of it.
    I suppose the question becomes: "just how often does it need to do this successfully?... what error rate are we willing to accept?" I would think that as long as it's a single tic better than the human rate, then it should be allowed on the roads. (Else, we should quit letting humans drive as well.)

    The question of where to set the bar ultimately comes down to where responsibility, and therefore liability is ascribed. A self driving vehicle is, ultimately a robot. If the AI controlling that robot is not recognized as a self-owned entity, then responsibility for the decisions it makes either come down to who/what owns the robot, or who designed and built said robot. Only once that distinction is clear will a consumer be able to make an informed decision as to whether his, her, or its risk aversion is high enough or low enough to delegate decisions to the AI or not. If you hire a chauffeur and said chauffeur makes a decision that jeopardizes persons or property, the chauffeur is ultimately the bearer of responsibility; however, unlike the chauffeur, as long as the AI is not recognized widely, if not universally, as its own entity, then either the robot's owner, or the robot's designer & builder will shoulder the liability for harm alleged to be caused by the AI. The very fact of a lack of consensus over where liability should lie causes uncertainty for potential consumers. This uncertainty serves to drive the acceptable level upward, to act as a hedge. The less clear who or what bears responsibility for harm to persons or property, the more certainty consumers will seek that there won't be a need to identify who or what is liable.

    For example, if I, as consumer, am considering a roving robot to transport goods or passengers for my business, and the manufacturer agrees to fully indemnify me and my business against all possible liability, then my risk is lower and I can be more tolerant of a lack of certainty of the capabilities of the AI. If, on the other hand, I agree to bear full risk for the decisions of the AI, then I want to be as close to absolutely certain that nothing will go wrong, no matter how bizarre or challenging the situation. If I don't know where in the continuum between those two extremes liability and responsibility falls, that uncertainty may make me even less tolerant of risk, to the point where I want something with slick weather vehicle handling skills that would put a Group B rally driver and ice racer to shame, regardless of how absurdly high an unreasonable that standard may be.

    This isn't merely a question of the capabilities of an AI & its underlying hardware and software.

    muthuk_vanalingam
     1Like 0Dislikes 0Informatives
  • Reply 13 of 14
    hentaiboyhentaiboy Posts: 1,252member
Sign In or Register to comment.