Slashdot Mirror

← Back to Stories (view on slashdot.org)

MIT Invented a Tool That Allows Driverless Cars To Navigate Rural Roads Without a Map (vice.com)

Posted by BeauHD on from the off-the-radar dept.

An anonymous reader quotes a report from Motherboard: A student at MIT's Computer Science and Artificial Intelligence Lab (CSAIL) is developing new technology, called MapLite, that eliminates the need for maps in self-driving car technology altogether. This could more easily enable a fleet-sharing model that connects carless rural residents and would facilitate intercity trips that run through rural areas. In a paper posted online on May 7 by CSAIL and project partner Toyota, 30-year-old PhD candidate Teddy Ort -- along with co-authors Liam Paull and Daniela Rus -- detail how using LIDAR (a radar-like sensor that uses lasers instead of radio waves to measure distances) and GPS together can enable self-driving cars to navigate on rural roads without having a detailed map to guide them. The team was able to drive down a number of unpaved roads in rural Massachusetts and reliably scan the road for curves and obstacles up to 100 feet ahead, according to the paper.

69 comments

  1. Driver by Anonymous Coward · · Score: 0

    Is it called?

    1. Re:Driver by BeauHD+(4) · · Score: 0

      It is called Progress.

      MIT is *the* one and only collage that give's UC Berkeley a run for its money.

    2. Re: Driver by Anonymous Coward · · Score: 0

      MIT is *the* one and only collage that give's UC Berkeley a run for its money.

      In the US. Big *the* it is not.

    3. Re:Driver by Z00L00K · · Score: 1

      I'm more worried about the cases on rural roads where the surface sometimes is soft and cars will get stuck severely on the road trying to drive where it shouldn't.

      --
      If builders built buildings the way programmers wrote programs, then the first woodpecker would destroy civilization.
  2. Big deal - so did Uber by 93+Escort+Wagon · · Score: 2

    When you eliminate the requirement to avoid running into things, the problem gets a LOT simpler!

    --
    #DeleteChrome
    1. Re:Big deal - so did Uber by Anonymous Coward · · Score: 0

      Lawyers will sue this into oblivion poste haste

    2. Re:Big deal - so did Uber by Anonymous Coward · · Score: 0

      I can do this with two cameras on my head. Millions do it every day like this. Two cameras do not give you much in terms of stereoscopic vision moreso than one camera does for distances involved with driving.

      You don't need a $10,000 lidar. All you need is cheap cellphone cameras and the right software.

      Comma.ai is already doing this. It's open source, check it out!

    3. Re:Big deal - so did Uber by Anonymous Coward · · Score: 0

      And why just rural roads, wouldn't the same tech be useful in town??

    4. Re:Big deal - so did Uber by Anonymous Coward · · Score: 0

      93 Escort Wagon opined:

      When you eliminate the requirement to avoid running into things, the problem gets a LOT simpler!

      Thanks for clarifying the fact that you have no meaningful, personal experience driving on rural (for which read "dirt") roads.

      The thing is - and I speak from some years' experience living on such a road - they are absolutely not dependably free of obstructions. In fact, probably the most common obstruction drivers run into (a phrase I mean literally here) is deer.

      See, deer routinely behave in what can best be described as "suicidal" ways. (I choose not to say "irresponsible ways," instead, because anyone who expects a wild animal to behave responsibly from a human perspective should not be permitted to cross the street without close, adult supervision - for his or her own protection.) They are infamous for leaping out into the path of oncoming drivers so abruptly - and often from within thick, concealing, roadside vegetation - that it's literally impossible for the human to react in time to avoid colliding with them (because there's a hard, physiological limit on the round-trip speed of nerve impulse transmission from the eye to the brain and back out to the motor nerves of about a half second).

      No matter how attentive you are to road conditions and the effects of weather and ambient light levels on visibility, the damned deer will manage, without warning, to place themselves squarely in your path, in close enough physical proximity to guarantee you will hit them. Rural residents live and drive in mortal terror of the fucking things - and that fear is absolutely justified ...

      (Posting as AC only so as not to undo prior upmods in this thread.)

      --

      Check out my novel ...

    5. Re:Big deal - so did Uber by Immerman · · Score: 1

      Actually it's only about 1/8th of a second of lag as I recall - at least sending signals to the hand, might take slightly longer to the foot. Classic example - have someone rest their hand on a table, extending their fingers past the edge, with thumb and forefinger poised to grab the midpoint of a dollar bill as soon as it's dropped. It's physically impossible to do so as the bill takes ~1/10th of a second to fall half it's length, and thus is safely clear before the signal gets from eye to brain and then to finger tips.

      Of course that's with the subject poised to act in a predetermined fashion the moment a signal is given. Take them by surprise instead and things no doubt slow down a lot.

      --
      --- Most topics have many sides worth arguing, allow me to take one opposite you.
  3. good movie idea by john+of+sparta · · Score: 2

    maybe...Texas MapLite Massacre

  4. As someone.... by GerryGilmore · · Score: 1

    ...who both lives in a rural are AND has been an advocate for this kind of technology long before it was remotely feasible (mid-1980s), I am excited about any advance in this realm.
    I've lost way too damned many friends and fam to "bitterly cling" to my self-driving car.
    And I say this as a proud former owner of a 1970 Ford Torino Cobra with the 429 CJ engine, shaker hood and rear window louvers (came THIS close to getting a Boss 429 Mustang), 4-speed manual, etc. Fun times.
    Things are different today, though....

    1. Re: As someone.... by Type44Q · · Score: 1

      ...AND has been an advocate for this kind of technology long before it was remotely feasible

      So you're a moron... AND it's okay because... why, again??

    2. Re:As someone.... by Anonymous Coward · · Score: 0

      >(came THIS close to getting a Boss 429 Mustang)

      You fucked up

    3. Re: As someone.... by Anonymous Coward · · Score: 0

      "I've lost way too damned many friends and fam to "bitterly cling" to my self-driving car." ...do what now? How rich are you guys that you all have had self-driving cars long enough to lose friends and family?

    4. Re: As someone.... by Anonymous Coward · · Score: 0
      Don't worry, as he said:

      I've lost way too damned many friends and fam to "bitterly cling" to my self-driving car.

      The Darwin principle is in full effect here.

    5. Re: As someone.... by GerryGilmore · · Score: 1

      Oh, I don't know...let's start with: saving lives and debilitating injuries; allowing homebound people who can no longer drive to (safely!!) get to the grocer, doctor, etc. You know - moron kind of stuff. (Fucking Idiot!!)

    6. Re: As someone.... by 110010001000 · · Score: 2

      I'll never get why people are so hung up on self-driving cars. Either drive yourself or hire a taxi. We should focus on making computers do stuff that we can't do ourselves.

    7. Re: As someone.... by Wycliffe · · Score: 3, Insightful

      Pretty much everything a computer does we can do. We program them to do it. There are plenty of things that computers can do faster and better though. Hopefully self driving cars becomes on of those. If you donâ(TM)t think self driving cars are worth it, you arenâ(TM)t thinking big enough. Imagine taking a cross country road trip where each morning you wake up at a different national park. Thatâ(TM)s the type of thing that is possible once we have self driving cars. Not to mention the price of a taxi drops considerably as does the cost of shipped goods. There are likely tons of spin off technologies we havenâ(TM)t even imagined yet that will become possible and cost effective once we have self driving vehicles.

    8. Re: As someone.... by Anonymous Coward · · Score: 0

      "We program them to do it."

      Who is we, you? Are you the jerkoff who told the Uber car to just "run into the big thing in the lane ahead of you". The programmers --- oh excuse me --- engineers who programed that car should be held for vehicular homicide. The car did just what "they" told it to do ---- hit the big thing in the road ahead.

    9. Re:As someone.... by HornWumpus · · Score: 0

      Things are different today, though....

      Whoever has your balls, should at least give you visitation.

      --
      John McAfee 'It was like that time I hired that Bangkok prostitute; to do my taxes, while I fucked my accountant'
    10. Re: As someone.... by Kjella · · Score: 1

      I'll never get why people are so hung up on self-driving cars. Either drive yourself or hire a taxi. We should focus on making computers do stuff that we can't do ourselves.

      I know how to do addition, but if you ask me to tally up a million records it's going to take a while and have errors. So there's plenty reason to make computers do what we do if they do it better. Sure replacing car drivers with self-driving cars might seem a bit pedestrian, like replacing a bunch of people with calculators with a computer. But there's more than a billion of them. Sure you can find lots of things to automate that affect a thousand people. Some that affect a million people. But there's very, very few things that would affect a billion people.

      That goes for jobs too, sure you can divide people into sectors but reality is most do their own fairly unique job, flipping a burger isn't the same as making fries to a robot. But driving a car is something lots and lots of people do for a living, all on the same roads playing by the same rules. If you can pass all the practical and legal hurdles the "smart car" could be as big as the smartphone. Maybe even bigger. Last I heard that was a fairly profitable adventure for Apple...

      --
      Live today, because you never know what tomorrow brings
    11. Re: As someone.... by Anonymous Coward · · Score: 1

      > I'll never get why people are so hung up on self-driving cars. Either drive yourself or hire a taxi. We should focus on making computers do stuff that we can't do ourselves.

      Like driving?

      I have a lot of older aunts and uncles who are quickly losing their ability to safely drive themselves to doctors appointments, grocery shopping, etc, and either live in a village so small that there's no taxi around, or they simply can't afford the service. Their (adult) kids' work schedules don't necessarily allow them to take time off to look after their parents. I can certainly see some of these folks come to rely on self-driving cars when they finally become reliable - and they can't come fast enough.

    12. Re: As someone.... by Anonymous Coward · · Score: 0

      "programmers --- oh excuse me --- engineers"

      You keep using that word. I don't think it means what you think it means.

    13. Re: As someone.... by Toshito · · Score: 1

      Imagine taking a cross country road trip where each morning you wake up at a different national park.

      That would be a pretty boring trip.

      Part of the fun of doing a car trip is having to navigate, finding unexpected things en route and stopping to see them, etc.

      Driving a car is so much fun, why would you want to miss that and sleep instead?

      --
      Try it! Library of Babel
  5. Missing the word detailed by Anonymous Coward · · Score: 0

    It still uses maps, just not detailed maps.
    From the abstract "In this paper, we address the problem of autonomous navigation in rural environments through a novel mapless driving framework that combines sparse topological maps for global navigation with a sensor-based perception system for local navigation."

  6. lasers instead of radio waves by fahrbot-bot · · Score: 1

    ... LIDAR (a radar-like sensor that uses lasers instead of radio waves to measure distances) ...

    (sigh)

    --
    It must have been something you assimilated. . . .
    1. Re: lasers instead of radio waves by Anonymous Coward · · Score: 0

      Obviously they got confused by the L in the acronym.
      But how is a normal human reporter supposed to know they actually use a disco ball?

    2. Re:lasers instead of radio waves by FatdogHaiku · · Score: 1

      Remember back when every simple term didn't have to be explained in the summary?
      Like LIDAR is cutting edge or something.
      Come on, it's not 3D printed rocket science...
      (but that may come up soon)

      --
      You have the right to remain sentient. If you give up the right to remain sentient, you will be elected to public office
    3. Re:lasers instead of radio waves by Immerman · · Score: 3, Informative

      > it's not 3D printed rocket science...(but that may come up soon)

      Actually it's been going up for just under a decade, longer if you don't insist on reaching orbit. The Falcon 1 Flight 4 reached orbit September 28, 2008 using its 3D printed engine components.

      --
      --- Most topics have many sides worth arguing, allow me to take one opposite you.
  7. This is good by PPH · · Score: 1

    I'm just surprised that this kind of navigational problem wasn't made a part of the baseline autonomous car requirements. You might think that unmarked, poorly mapped roads are part of a rural setting. But we've go crap like that right in town. And self driving cars are going to have to deal with it.

    --
    Have gnu, will travel.
    1. Re:This is good by 110010001000 · · Score: 2

      I am not surprised. Autonomous driving is a joke. Sure you can build a system to get about 80% there as long as it has high resolution maps.

    2. Re:This is good by Anonymous Coward · · Score: 0

      Except autonomous driving is 99.9999% there already.

      Slashdot Luddites, however, are still a joke. If only we could fix them.

    3. Re:This is good by vtcodger · · Score: 2

      "I'm just surprised that this kind of navigational problem wasn't made a part of the baseline autonomous car requirements."

      Me too. I don't see how an autonomous vehicle can handle construction zones, parking garages, or GPS "dead zones" without this sort of technology.

      --
      You can't see ANYTHING from a car, You've got to get out of the goddamned contraption and walk...Edward Abbey
    4. Re:This is good by Anonymous Coward · · Score: 0

      Autonomous driving is a joke.

      Slashdots posts are a joke.

      It was a time since we reached a point where some companies have pretty good algorithms that they don't take on the road because they feel they aren't ready yet and other companies thought it would be a good idea to experiment with code they know is flawed in live traffic.

      Please don't push for a culture that generalizes and removes the fun from people who behaves just because it is easier than sort of those who misbehaves.

    5. Re:This is good by drinkypoo · · Score: 2

      Me too. I don't see how an autonomous vehicle can handle construction zones, parking garages, or GPS "dead zones" without this sort of technology.

      To handle construction zones, a level 5 vehicle will need advanced information about the construction zone. That means that construction workers are going to have to place beacons. That will become the law sooner or later. A level 4 or lesser vehicle will just expect you to take over.

      As for the rest, a level 5 vehicle will simply refuse to go anywhere it doesn't understand, while again, a level 4 or lesser vehicle will expect you to take over. A level 4 vehicle should give you lots of notice.

      --
      "You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
    6. Re:This is good by PPH · · Score: 1

      That means that construction workers are going to have to place beacons.

      The power to the hospital is out. A tree fell across the lines.

      "Sorry. But we don't have our beacons with us. And we can't push an update to The Cloud. Because a tree fell across the lines and communications are down. You'll just have to wait."

      a level 5 vehicle will simply refuse to go anywhere it doesn't understand

      Suits me. Just keep those shitboxes off the roads. And good luck surviving through even a minor storm when a bit of high water or some (otherwise navigable) debris falling in the roadway keeps people housebound and starving to death.

      --
      Have gnu, will travel.
  8. Significant improvement. by Gravis+Zero · · Score: 1

    This is a step in the right direction. Instead of relying on ultrahigh precision LIDAR maps, they now rely on basic map data your GPS nav unit would have + LIDAR for local information. What this really means is that the level of reliance on LIDAR has dropped significantly. Elon musk called LIDAR a crutch and this, for the most part, ditches that crutch because it's capable of operating in an "unstructured area" (place that hasn't previously been mapped) by using it's "local" sensor (LIDAR in this case) to determine the edges of the road. It's now a matter of implementing a similar sensing system using RADAR and computer vision and your Tesla can autonomously drive you around, disregard and run into pedestrians too! ;)

    link to the the paper describing the system.

    --
    Anons need not reply. Questions end with a question mark.
    1. Re:Significant improvement. by 110010001000 · · Score: 1

      I don't get it. I would imagine it senses the edges of the road (probably based on texture differences between the road and the grass) and keeps the car in the road. How is this significant? It also uses openstreetmap data to help with the navigation prediction (I guess).

    2. Re:Significant improvement. by Gravis+Zero · · Score: 1

      It previously relied on models generated using detailed point cloud data. The real show stopper will be when it can recognize gravel and dirt roads that aren't on any maps.

      --
      Anons need not reply. Questions end with a question mark.
    3. Re:Significant improvement. by Anonymous Coward · · Score: 0

      Don't see how that's a "show stopper." Our vehicles can already do that.

      Reliance on LIDAR hasn't decreased at the vehicle level. RADAR and cameras are nowhere near as accurate as LIDAR. Tesla autopilot can barely operate on-road without it, no way I would trust it out in the boonies on a dirt road.

  9. MIT braggarts by Anonymous Coward · · Score: 0

    MIT is so overmarketed it is ridiculous.

    Everyone is doing the same experiments with the same tradeoffs. They showed up and had a go. Impressive, because there's a barrier to entry, but to take what's basically an experiment, tuning tools that already exist, and brand it "MapLite" is ridiculous. Being able to run without a map is not a binary feature. It's something you get better and better at doing. They do not have a system good enough to run without a map. Nobody has a system good enough to run period, with or without a map, or it would already be for sale.

    This is ridiculous blow-hard posturing on its face. They should dial it back or they will not be taken seriously.

    1. Re:MIT braggarts by mrclevesque · · Score: 1

      Yes. It seems they don't have much more than a flatness detector algorithm to point the car to where the 'road' is in rural areas.

  10. Yeah, it works great! by Anonymous Coward · · Score: 0

    In a laboratory. Nothing to see here.

  11. This is not new by neoRUR · · Score: 4, Informative

    None of this work is new, CMU and others were doing this 20 years ago. And we were using it on the Unmanned Ground Vehicles Project. We were not using GPS, but Neural Network Road following outdoors without roads with LIDAR and with cooperative robotic HUMVEE vehicles.

    1. Re:This is not new by Anonymous Coward · · Score: 0

      I was looking for this comment. I fail to see how/why this is news.

    2. Re:This is not new by Anonymous Coward · · Score: 0

      Yes, and 20 years ago CPU was not great - um 386?. Now consider a Bee, Fruit-fly or an Ant. Pigeon? Not much upstairs, but enough to get from a to b.
      Someone has overthought the problem. I figure thermal radiation differences. Plus more for multilane roads (white paint is not much, signs, or indicators).
      Poor old car now needs to film driver and log sensors. The real key is to program all those exceptions.

    3. Re:This is not new by Anonymous Coward · · Score: 0

      I fail

      Could have stopped right there.

    4. Re:This is not new by Anonymous Coward · · Score: 0

      None of this work is new

      Except for the GPS integration, topological data from OSM, and being designed for passenger cars on roads vs. military vehicles and 20-odd years worth of technological improvements. But other than that, yeah, the CMU and MIT projects are practically identical.

  12. Any Real test of driverless cars? by Anonymous Coward · · Score: 0

    Having grown up in Wyoming I have a real world test for these driverless cars to undertake that will get nearly everyone's attention.

    Four inches of snow over hard packed snow and Ice. Twenty Five mph (40 kph) winds producing ground blizzards. At night in -20 F (-28 C). Under these conditions, the snow will stick to many things on the down wind side.

    Or during a snow storm with about 4 inches of power covering the ground. The snow coming down looks like a Windows 3.0 screen saver but with more objects and they move around more. At night again.

    Or during a downpour from a super cell at night where the rain appears like it is in a strobe light (2016 Challenger in South Dakota). That one almost hurt.

    Show me a car that can drive under those conditions and I will take notice. Arizona and Southern California are the best case scenarios for driverless cars as they really don't get weird (or in the case of WY and SD "normal") weather. What do you want to be that when it starts to rain heavy in SCA that the driverless cars are in the garage. If Uber's car can't detect a single person crossing the street - what would they do with a few million moving targets? Pull over and cry?

    1. Re:Any Real test of driverless cars? by HornWumpus · · Score: 2

      I've been in downpours/hail that had everybody pulled over and hiding under overpasses. I'd be impressed if the autonomous car had the sense.

      --
      John McAfee 'It was like that time I hired that Bangkok prostitute; to do my taxes, while I fucked my accountant'
    2. Re:Any Real test of driverless cars? by Anonymous Coward · · Score: 0

      They've been testing in Michigan in the winter for YEARS now.

      Stop being a Slashdot Luddite idiot.

    3. Re:Any Real test of driverless cars? by vtcodger · · Score: 1

      Hmmm. I wonder it Uber vehicles can distinguish a pedestrian from a tumbleweed? Perhaps that explains mowing the former down.

      --
      You can't see ANYTHING from a car, You've got to get out of the goddamned contraption and walk...Edward Abbey
    4. Re:Any Real test of driverless cars? by Anonymous Coward · · Score: 0

      If it can't they clearly made the wrong choice.

      I've literally never encountered objects on the road that it was OK to drive over rather than drive around or wait for to go away.
      Slowing down/stopping for everything and alerting me that I might need to manually drive over something sounds like a perfectly viable solution for that once in a lifetime schroedingers cardboard box that may or may not contain a child is on the road.

      The focus should be less on trying to identify what is on the road and more on not running it over regardless of what it is.
      If it can't drive in a snowstorm or rain it is perfectly fine. I can take care of those times.
      I just want something that every now and then can take over that long boring section of the highway or the infuriating stuttering in a long car queue.
      The rest I don't need until I'm to old to drive.

    5. Re:Any Real test of driverless cars? by Anonymous Coward · · Score: 0

      The problem is, that if you drive less, your driving abilities will deteriorate over time so when you finally have to take over you will no longer being able to handle the car properly and that in a situation where you really need to be able to drive.

      Also LIDAR... So the car won't work in heavy rain or snowfall? That could you get stuck in the wrong place.

    6. Re:Any Real test of driverless cars? by Anonymous Coward · · Score: 0

      Don't matter if you can't drive yourself. If the car stops because there is a football in the road and you don't know how to drive over it - you go out and remove the football. Then you get in, and the automatic car resumes driving on the clear road.

      Why stop for a football? Because a cars limited intelligence can't distinguish it from a bowling ball, or a head poking out from a manhole.

    7. Re:Any Real test of driverless cars? by Immerman · · Score: 1

      Just don't forget to turn off the car before removing the football, or you may find it has resumed driving without you.

      --
      --- Most topics have many sides worth arguing, allow me to take one opposite you.
  13. navigate? without map? by 4wdloop · · Score: 2

    So...does it stop a lot to ask for directions?

    --
    4wdloop
    1. Re:navigate? without map? by Anonymous Coward · · Score: 0

      So...does it stop a lot to ask for directions?

      In Massachusetts you are virtually always within view of a Dunkin Donuts so you can use them as landmarks.

    2. Re:navigate? without map? by cascadingstylesheet · · Score: 2

      So...does it stop a lot to ask for directions?

      Maybe it keeps one hand on the left wall at all times ...

    3. Re:navigate? without map? by Anonymous Coward · · Score: 0

      that works great until you hit a roundabout...

  14. Lidar is limited by Anonymous Coward · · Score: 0

    Very nice development but it's also good to know that LIDAR has severe limitations when used in fog, rain, snow and dusty conditions. Maybe there are ways to work around this but I would be careful to let an autonomous driving system be totally reliant on LIDAR.

  15. 2005 Just Called by Anonymous Coward · · Score: 0

    They wanted to remind you that Stanford won the DARPA Grand Challenge with a LIDAR-equipped vehicle back in 2005. Is MIT now claiming that they independently invented this only 13 years later?

  16. Oncoming Traffic by John.Banister · · Score: 1

    I hope they teach them to pull over at the wide spot and wait when they see oncoming traffic on a 1.5 lane road. Flashing their lights to alert oncoming drivers of new hazards would also be nice.

  17. Wake me when it can do this by RogueWarrior65 · · Score: 1

    https://en.wikipedia.org/wiki/...

  18. And when it really gets lost.... by Anonymous Coward · · Score: 0

    It will stop and ask the locals for directions. That is of course the manufacturer uses female programmers.

  19. So MIT invented the horse? by Pyramid · · Score: 1

    They must party like it's 1899 over there...

    --
    ~Any apparent grammatical or typographic errors are caused by defects in your display device.
  20. Needs to be further than 100 feet. by Anonymous Coward · · Score: 0

    Needs to be well beyond the stopping distance of the vehicle.

  21. Oh well.. by Anonymous Coward · · Score: 0

    There go the cows..