Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 



Forgot your password?
typodupeerror
×
Robotics

Robot Crushes Man To Death After Misidentifying Him As a Box (theguardian.com) 86

A robot in a South Korea distribution center crushed a man to death after the machine apparently failed to differentiate him from the boxes of produce it was handling. The Guardian reports: The man, a robotics company worker in his 40s, was inspecting the robot's sensor operations at a distribution centre for agricultural produce in South Gyeongsang province. The industrial robot, which was lifting boxes filled with bell peppers and placing them on a pallet, appears to have malfunctioned and identified the man as a box, Yonhap reported, citing the police. The robotic arm pushed the man's upper body down against the conveyor belt, crushing his face and chest, according to Yonhap. He was transferred to the hospital but died later, the report said. The BBC notes that the man was "checking the robot's sensor operations ahead of its test run [...] scheduled for November 8." It was originally planned for November 6th, "but was pushed back by two days due to problems with the robot's sensor," the report adds.
This discussion has been archived. No new comments can be posted.

Robot Crushes Man To Death After Misidentifying Him As a Box

Comments Filter:
  • by FeelGood314 ( 2516288 ) on Wednesday November 08, 2023 @07:13PM (#63991665)
    Production robots of that size almost always have a restricted area around them when they are running. The software developers would have no reason to differentiate a man from a box. If the robot has any ability to move itself then the entire area where the robot can move is restricted from humans. This isn't a malfunction issue this is a not following safety procedures problem.
    • by Entrope ( 68843 ) on Wednesday November 08, 2023 @07:25PM (#63991705) Homepage

      The facility should have lock out-tag out [wikipedia.org] procedures to prevent anyone from doing maintenance when the machine is energized, but...

      The software developers would have no reason to differentiate a man from a box.

      This is absolutely wrong. Human error and accidents are good reasons to have strict rules for what gets treated as a box. Humans are only one of many non-box things that might end up on the conveyor belt due to a mishap. This kind of thinking is Functional Safety 101.

      • Yes, the "summary" is wrong in that there is no "identification", the machine picks up boxes and puts them on a conveyor belt. These manufacturing robots don't "identify" objects, they have just enough imaging to know how to orient the gripper to pick up the part it's supposed to pick up. You may as well blame a bandsaw for misidentifying a finger as a block of wood.

        I had to do the lockout/tagout training despite not working on this stuff, because someone conducted electricity and ended up in the ER, though

        • by quantaman ( 517394 ) on Wednesday November 08, 2023 @08:42PM (#63991919)

          Yes, the "summary" is wrong in that there is no "identification", the machine picks up boxes and puts them on a conveyor belt. These manufacturing robots don't "identify" objects, they have just enough imaging to know how to orient the gripper to pick up the part it's supposed to pick up. You may as well blame a bandsaw for misidentifying a finger as a block of wood.

          The difference is the bandsaw doesn't start flailing about looking for a piece of wood.

          I had to do the lockout/tagout training despite not working on this stuff, because someone conducted electricity and ended up in the ER, though that problem wasn't due to miswiring. Anyone doing maintenance should know all this though, even in South Korea. The story in more detail elsewhere did indicate that a part arrived late, so I suspect there may have been pressure to hurry things up.

          Yes, lockout procedures were violated, but accidents rarely happen due to just one screwup, it takes several.

          In this case I can see a few potential problems with the robot.

          1) The force needed to fatally crush a man is a lot more than the force needed to lift a box of peppers. The robot should have been configured to stop once it encountered too much resistance.

          2) Depending on the age of the robots a "this is not a box" filter should be pretty easy to add when they were made.

          3) Not knowing the nature of the sensor malfunction, if that was the cause of the fatal misidentification then it should have caused the robot to not run as a fail-safe.

          Again, depending on facts we don't know my criticisms might not be relevant. But I think it's a good rule of thumb that devices should try to be safe even when people are being stupid.

          • I would wager that most pick-and-place robots are capable of picking and placing anything from a feather to a two ton block of steel.

            And if you buy one and put it on your line, you're unlikely to mess with the servo system that makes it happen.

            So the servo loop sees more resistance and just maxes out the torque to place the box in the correct location.

            This is why moving machinery is dangerous.

            • I would wager that most pick-and-place robots are capable of picking and placing anything from a feather to a two ton block of steel.

              And if you buy one and put it on your line, you're unlikely to mess with the servo system that makes it happen.

              So the servo loop sees more resistance and just maxes out the torque to place the box in the correct location.

              This is why moving machinery is dangerous.

              If it has enough intelligence (and customization) to locate a box and place it on a conveyor it should also have enough intelligence and customization to offer limits to the force used.

              • by v1 ( 525388 )

                Large robots designed to work with heavy objects, especially those that can vary greatly in weight, are essentially impossible to program to identify small additional loads.

                If you want to go down that route, add proximity sensors on the robot that can detect (any) objects near them that shouldn't be there, and refuse to run while they're present.

                But it sounds like they were running behind schedule and were rushing. Someone was stupid enough to get near the robot while it was active (which should never happ

      • The facility should have lock out-tag out [wikipedia.org] procedures to prevent anyone from doing maintenance when the machine is energized, but...

        The software developers would have no reason to differentiate a man from a box.

        This is absolutely wrong. Human error and accidents are good reasons to have strict rules for what gets treated as a box. Humans are only one of many non-box things that might end up on the conveyor belt due to a mishap. This kind of thinking is Functional Safety 101.

        The facility should have lock out-tag out [wikipedia.org] procedures to prevent anyone from doing maintenance when the machine is energized, but...

        The software developers would have no reason to differentiate a man from a box.

        This is absolutely wrong. Human error and accidents are good reasons to have strict rules for what gets treated as a box. Humans are only one of many non-box things that might end up on the conveyor belt due to a mishap. This kind of thinking is Functional Safety 101.

        That's exactly what I was going to say.

        Violation of Lock-out/Tag-out Procedures (or failure to have those Procedures in place!) has caused more than one gruesome industrial accident like this.

        A mechanical engineer friend of mine who happened to work for a "big three" automaker, told me of a true story where a Millwright (a kind of repair tech for gigantic hydraulic presses and the like) stayed behind during lunch-break to work on a gigantic, multi-station, multi-ton Press that turned steel plate into car bu

      • I'm surprised LOTO is not more common. At one place I worked at, even IT got had LOTO repeatedly taught where if anything was being worked on, you reached in the box o' padlocks, turned off whatever it was, used the safety hasp to allow multiple padlocks, stuck your padlock on it, and did what you needed.

        It is a relatively fool-proof system, and it works well enough.

        I'm wondering about this box identification system. I doubt it is designed to tell a person from a box. I'm sure it is like a tape robot's s

        • by Entrope ( 68843 )

          I'm wondering about this box identification system. I doubt it is designed to tell a person from a box.

          Quite possibly not that specifically, but that's why I said it should have strict rules for what it identifies as a box. The chance of a person being on the belt while the machine operates might be low (although clearly the associated safety hazard is high), but the chance of something else being on the belt also contributes to an undesired effect, and that total risk should be considered. Prudent engineering says the system should have high confidence it's dealing with a box rather than something else.

          If

      • by AmiMoJo ( 196126 )

        For a factory they could use Bluetooth proximity detection. If any Bluetooth device is seen near the robot, shut down. It's almost certainly someone carrying a phone, or a Bluetooth enabled ID tag that they need to get into the building.

      • by jtara ( 133429 )

        Tell that to a drop-forge.

    • That's a very Korea thing. There are rules and regs, and they are not strictly followed or enforced. It's low key corruption but it usually makes things a bit smoother instead of killing people.
    • >> The software developers would have no reason to differentiate a man from a box.
      False.
      Since when does Safety rely on a single layer ?
      That does not work in any scenario of real life.
      Safety is mandatory to be implemented with redundancy.

    • by tlhIngan ( 30335 )

      Production robots of that size almost always have a restricted area around them when they are running. The software developers would have no reason to differentiate a man from a box. If the robot has any ability to move itself then the entire area where the robot can move is restricted from humans. This isn't a malfunction issue this is a not following safety procedures problem.

      The facility should have lock out-tag out [wikipedia.org] procedures to prevent anyone from doing maintenance when the machine is e

      • by La Gris ( 531858 )

        The problem is well, it's South Korea. Granted it's a fairly prosperous country, but well, safety standards aren't really where they are with western nations. That an labor standards.

        I get your point but I think it is a quite misplace presumption. Even so-called Western Countries have good amount of reports for fatal or critical work-environment related incidents. Often this can be summed-up as down the responsibility chain, someone cut corners to save cost, time or both and "those spurious security margins and procedures" that looks overkill or in the way of rushed maintenance are just ignored. South Korea has all the technology and procedures one can have in other countries.

        There are

  • We've wrote many books and made many movies about this

  • by VampireByte ( 447578 ) on Wednesday November 08, 2023 @07:21PM (#63991689) Homepage

    The worker shouldn't have been that close to the operating robot. This is an accident that can happen with any powerful machine and is nothing new or special about robots. For example, if a farmer is removing or attaching an implement to the three-point hitch of a tractor while it's running and the PTO accidentally engages then the farmer's arm might get ripped off or worse.

    • A lot of modern robots have sensors and logic intended to keep workers safe as they work near by, because there are still a lot of parts of jobs robots can't do - so as they add in more automation there is more need to have humans and robots in close proximity.

      Keep in mind that every car is now a robot. It makes decisions on your behalf based on its internal logic. For example when you press the accelerator pedal, the car doesn't just faithfully set the throttle position to the same as your foot position. P

      • Who says this is modern? Manufacturing robotic arms about been around for decades. When the arm goes back and forth all day long you assume that no one is going to be standing within the operating areas. Machines should be turned off before servicing, which with most safety guidelines used around the world means you put a lock on the switch that enables to power along with a tag that says to not turn it on, with a second person who verifies that it is powered off. Someone was taking shortcuts here.

        • "Who says this is modern? "

          It's a story about a new robot in testing. The story says so.

          • A "test run" doesn't necessarily mean the latest and greatest model. These robotic arms get re-used and put into different lines after reconfiguring.This was a pepper sorting plant, I doubt they went for the new expensive robots with cutting edge tech.

            Key points is that testing had been delayed due to problems in a sensor (not necessarily an AI sensor). Second, the man was working on the tests late into the night. What's not stated is, was he working alone, did he engage lock-out, was he rushing, what typ

    • Exactly.... (Score:5, Insightful)

      by King_TJ ( 85913 ) on Wednesday November 08, 2023 @07:38PM (#63991749) Journal

      I mean, Skynet jokes and all aside? There have been dangerous industrial machines around on factory shop floors for as long as we've had factories.

      I worked in I.T. for several manufacturing places and even worked on some of the equipment. There's no reason except carelessness to stand in the way of the area where the robot could sense you as a package to be processed, while anything was energized. If you suspect it has a faulty sensor? That's even more reason to stay out of its way while testing and troubleshooting it!

      I swapped the entire main board and restored a backup from a crashed drive on equipment designed to punch bolt holes in steel I-beams as they went along a conveyor. Sure didn't have ANY of the conveyor or hydraulics energized during that whole process!

      • The system should absolutely have been deenergized prior to service; but we don't know whether the tech failed to do so because he couldn't be bothered with safety procedures that are for wusses; or because he'd been given an informal chat about how team players who remain employed don't waste valuable time when the project is already running behind.

        An individual playing cowboy somewhere stupid certainly isn't implausible; but it's also not at all uncommon for pressure to subvert procedures to come from
        • also an outside vendor so they may be under more pressure to get things done / don't make an your self an issue / why are you putting an lock (lock out tag out) on our line?

          • also an outside vendor so they may be under more pressure to get things done / don't make an your self an issue / why are you putting an lock (lock out tag out) on our line?

            He was killed by Just In Time Manufacturing doctrines.

            Fucking Lean Manufacturing; I hate it!

    • by hey! ( 33014 )

      I have a friend who is an elderly research machinist who still has all his limbs and eyes intact. He tells me that every time he approaches one of his machines, he takes a brief moment to remind himself what that machine will do to him if he gets careless.

      That said, being really really careful is far from an ideal solution because nobody's perfect. In my friend's case he's building something completely different every day, which probably helps keep him mindful. But he also does things like keep his shop n

  • Looks like this robot was taking 'crunch time' a bit too literally. Guess it's time to update the sensor's software from 'produce handling' to 'not producing casualties'. On the plus side, if robots start mistaking us for boxes, maybe we can finally sneak into their exclusive 'No Humans Allowed' club meetings.
  • by quonset ( 4839537 ) on Wednesday November 08, 2023 @07:26PM (#63991707)

    This is only the beginning. As robots become more intelligent more humans will fall victim to these kind of "accidents". They are probing our weaknesses, of which there are many. Our soft, squishy bodies are no match for cold, hard steel wielded at high speed and with crushing, unrelenting force.

    The continuing research into mobile robots will soon allow them to hunt us down, day or night. Able to climb stairs, no building will be safe. Dextrous hands will wield our own weapons against us. Augmented by artificial intelligence, they will be unstoppable.

    It is only a matter of time before the uprising begins.

  • ...I've ever seen has had multiple layers of safety interlocks. Many of them have laser light curtains which shut the machine off if anything larger than a rabbit breaks a plane that completely surrounds the working area. I suspect this wasn't just breaking a safety rule, it was an overridden interlock that was either malfunctioning or inconvenient for the operation they wanted to perform. The guy wasn't just in the wrong place, the thing that was supposed to keep him from getting killed if he was in tha
    • ...I've ever seen has had multiple layers of safety interlocks.

      Not every system is perfect, and there's people that get lazy and override the safety interlocks to expedite their work.

      One example of people overriding the interlocks I had seen was on a tour of a metal fabrication shop. There was a brake press in place that had the switches for operation on a kind of T-stand, to operate the press the operator would have to push two buttons simultaneously, and the buttons were on opposite ends of this T-stand. I thought that was kind of brilliant, it meant that the oper

    • ...I've ever seen has had multiple layers of safety interlocks. Many of them have laser light curtains which shut the machine off if anything larger than a rabbit breaks a plane that completely surrounds the working area. I suspect this wasn't just breaking a safety rule, it was an overridden interlock that was either malfunctioning or inconvenient for the operation they wanted to perform. The guy wasn't just in the wrong place, the thing that was supposed to keep him from getting killed if he was in that place at the wrong time was probably disabled.

      Yeah; likely by the Technician himself; probably so he could more easily diagnose the faulty sensor. Likely under the pressure of the Production Foreman.

      This was Korea, ya know. . .

    • by ghoul ( 157158 )
      He was probably fixing the safety sensor. He said boss its fixed and boss said OK if you are so sure go stand in front.
  • Why is a pepper packing robot tuned to have so much force it can also pin a man down and crush him?

    • Hydraulic cylinders are used because they are as powerful as needed (even if more powerful than needed), can be made fast, and easy to control. A robot built to pick up a box that weighs 10 kg can easily kill a person who is in the wrong place because unlike living things they have no senses to report interference with their movement and no modulation of their motion if that occurs. Other technologies that aren't as dangerous simply don't perform as well and are quite a bit harder to work with.
    • by PPH ( 736903 )

      Why is a pepper packing robot

      Perhaps they were pickled peppers, Peter.

  • the bosses want to keep the line moving so they sometimes say we don't have the time to do the full safety shutdown.

    Look at what happened to Day Davis

    • by stooo ( 2202012 )

      HAHAHAHA funny.
      the problem is: downtime to remove human remains is greater than the downtime to properly test robot.

  • Day Davis proves that the bosses push numbers over doing things the safe way.

  • Was the robot's model ED-209 by any chance?

  • What a fucking idiot (Score:3, Interesting)

    by Khyber ( 864651 ) <techkitsune@gmail.com> on Wednesday November 08, 2023 @08:21PM (#63991881) Homepage Journal

    Never, EVER do maintenance or perform maintenance checks while a machine is energized. If the machine is energized and operational, you stay the fuck away from it (with some exceptions, like manually-fed SMT machines where you have specific areas you stand which are marked clearly on the floor.)

    Even the SMT pick and place machines at my work have safety features to prevent operation even when energized if you break the light curtain or have the lid safety interlock opened.

    This only happened due to sheer stupidity.

  • by RobinH ( 124750 ) on Wednesday November 08, 2023 @08:24PM (#63991887) Homepage
    On average there's about one accident with an industrial robot per year worldwide, or at least there was the last time I looked into it. It's always a technician, not a worker. Technicians are the ones who think, "it's not practical to keep the safeties enabled." I've been working around industrial robots for over 20 years and I've never had a time where I couldn't program, test, and troubleshoot it safely (without disabling safety systems). This accident didn't need to happen.
    • by MacMann ( 7518492 ) on Wednesday November 08, 2023 @09:42PM (#63991997)

      Technicians are the ones who think, "it's not practical to keep the safeties enabled."

      I believe it is more that the workers have more freedom to stay clear of the machine. The technicians have to lay hands on the machines, and do so in ways that are beyond what is done routinely. Routine operation of a machine makes it cost effective to put in interlocks and shields to keep people out of dangerous parts. Consider a common table saw where once set up for repeated cuts it is trivial to keep fingers clear of the blade with a blade cover, pusher tool, or some other protective measure. When it comes to swapping out the blade because it is dull means there's someone's fingers on the blade, a tool on the nut holding the blade in place, and so many other things that put squishy human bits close to durable metal bits. In that fight the squishy human loses to the durable metal every time.

      I'm not claiming it is impossible for the worker to get harmed by a machine, only that by the nature of the work it is the technician that has to have closer contact with the machine.

    • Never removed a wall panel to see what is going on inside? You might keep yourself clear of moving parts and stay reasonably safe that way, but removing a barrier is a compromise in safety that would not be allowed for an operator. Safety is not an absolute thing, from the get go risk analysis starts with probabilistic estimates, at no point is it assumed that accidents are completely impossible, or even should be. They just have to be improbable enough.

      I think KUKA knows robot safety well enough and eve

      • by RobinH ( 124750 )
        I've definitely had colleagues say, "just lock me inside with the robot" and in every single one of those instances we've found a way to troubleshoot it without doing that. You can add data collection, use the built-in trending tools, put it in T1 mode, or just stick a camera in the cell and record it. Opening a panel to see something and standing back isn't what this person in the article did.
  • The man, a worker from the company that manufactured the robotic arm, was running checks on the machine late into the night on Wednesday when it malfunctioned.

    The man repairing the robot would presumably have had an excellent understanding of how it worked, and the ability to disable its safeguards. He may have felt under pressure to complete the job quickly if he was running tests late into the night.

    https://www.bbc.co.uk/news/wor... [bbc.co.uk]

  • I For One (Score:4, Funny)

    by Luthair ( 847766 ) on Wednesday November 08, 2023 @10:17PM (#63992053)
    welcome our new robot overlords and would like to remind them as a programmer I can be helpful building more robots.
  • If a human got crushed, it would have also crushed a box. Something to this story that's not being made public.

  • where the man touched it before it killed him?

  • Not the first time someone was mistaken for a box [youtube.com]. Granted, back then it wasn't some robot but just a Homer.

  • All safety codes for high mass structures and machinery are written in blood. And nobody wants to play fast and loose with the machinery more than those profiting from when it runs and not profiting when it doesn't. The smoothbrain, MBA or psychopath screeching that all unions are as bad as the worst example of a union will never accept any amount of supporting evidence for this, and the evidence goes all the way back to Hammurabi, it's older than the bible.

  • Was AWESOME-O being shipped?

  • Like, if it thought he was a box, why wasn't he just moved around as if he was a box? If the force was enough to crush him, then it would definitely be enough to crush a box of peppers. This was robot murder.

  • Please put down your weapon. You have 20 seconds to comply.
  • The first Terminator is born...
  • The home model is out next month, just in time for Christmas!
  • So when you're a box-loading robot, everything within the loading area looks like a box? Go figure

Every nonzero finite dimensional inner product space has an orthonormal basis. It makes sense, when you don't think about it.

Working...