The renewed focus on reliability is motivated by emerging applications. Imagine a wireless factory robot in a situation where a worker suddenly steps in front of it and the robot needs to make an immediate decision.
This example is a real WTF. I really hope nobody is planning on building safety-critical real-time systems on top of WiFi!
I don’t know about manufacturing environments but I deal with laboratories a lot, and I’m a bit baffled at how quickly lab operators have jumped on battery-operated wifi sensors for lab monitoring systems. I have like three room sensors attached to my EcoBee thermostat at home and I can barely be assed to change the batteries in those things, I cannot imagine dealing with batteries and connectivity troubleshooting for a building full of sensors whose reliable operation is often critical for regulatory compliance. Seems like the perfect application for PoE systems, to me
In industrial there are very few wireless systems unless they are either too remote from the CPU and aren’t safety sensitive. Safety is taken very seriously because any incident can mean injuries/death and ending up in the public eye. Any safety systems are hard wired because of reliability.
If your robot moves around, then it needs a wireless connection. And it doesn’t really get any more reliable than wifi. I’m certainly not going to outsource that to a Verizon cellular connection.
And even for things that can be wired - ethernet is far from reliable. Cables are easily damaged or simply unplugged.
Wifi can work really well, especially with high end networking gear (and not, for example, the wifi access point you get for free from Verizon).
Anything automated that could be a threat needs to have safeguards. Needing constant wifi to prevent death or injury is not an acceptable safeguard.
Consider consumer/professional drones. If they lose connection they have on board protocols to mitigate hazards. Even then they are still governed by laws to isolate then from people because even those safeguards aren’t good enough. Suggesting that a robot could completely rely on wifi is preposterous.
As someone using various wireless standards over over twenty years and in IT dealing with wifi instability on basically a daily basis. No.
Wifi is a series of compromises to be convenient. It’s “good enough” for most of those but generally and increasingly in newer standards, the compromise is to drop stability for things speed. You’ll see this to be the case in a lot of professional wifi gear that will transfer you to a lower standard if it sees weaker signals to improve stability.
To make that concrete, a problem with wifi in an office is an embarrassing “I’ll call back on my phone” but a factory floor that could be millions of dollars of downtime to restart an entire chain of machines. Hardened industrial wiring and connections is well established and wifi is just not at that level. The poorly formed example of the robot was trying to convey their intention to start addressing that level of hardening.
All that said, based on my experience reading ieee articles this is all exaggerated. in reality we’re probably just getting more stable video calls at higher bandwidths. Still a win for the help desk techs everywhere and people with a heavy wall making Netflix flaky.
As long as they have a delay counter which immediately shuts the robot down when it hasn’t been answered within a certain time period it shouldn’t pose much of a problem if it has an E stop. Just inconvenient when it keeps shutting down all day.
I work in autonomous vehicle engineering. That’s not even on the table for something we’d consider doing. But China is trying to enter the market hard, and I am less sure they wouldnt do that.
I really hope nobody is planning on building safety-critical real-time systems on top of WiFi!
Lol my company runs safety critical systems for autonomous mining equipment over wifi and have been doing it for 15 years. You “just” need to engineer it correctly to take into account an imperfect link. Just like you need to to with ethernet, because every link is an imperfect link when you’re dealing with safety critical systems.
Generally it’s ok. Wifi handovers between APs are the main issue, our wifi clients have handover times of about 70 milliseconds or so normally which is fine with the engineering controls in place. A shift’s worth of operation usually involves a thousand or more handovers with no issues.
Our machines are run in exclusion zones, and every entrance to that area has a twin barrier setup that creates a “no man’s land” between machine and personnel. The width of that is dependent on the approach speed of the machine and the worst case scenario timeout of the safety system controlling that area. So if you stand on one side of the barrier and the machine comes at you full tilt and somehow loses wifi just as it hits its side of the barrier, it’s still guaranteed to stop before it hits you.
And yes, we rarely get interference, seeing as we usually operate several hundred metres underground. But you’d be surprised to see just how many wifi clients are floating around down there, most mining operations are heavily networked these days.
I was imagining one AP, not lots of APs as fixed infrastructure. Something about being underground and safety critical made me assume the place wouldn’t look like an office building.
70 millisecond handover is pretty good, I assume 802.11r? If you control the clients then all is good and fixable, but if they’re (thunderclap) external vendor controlled (lights flicker) I imagine some will keep trying to hang onto old out-of-range APs for too long.
The area we cover is usually up to a kilometre or two long.
We just use good old 802.11g. Our wifi clients are Moxa industrial units and I’ve found that they’re pretty consistently compatible with all sorts of AP hardware that sites use. Sites with Cisco hardware and controllers give us the most grief, Cisco loves Cisco and nobody else, of course.
As wifi is pretty much line of sight underground (it doesn’t penetrate rock), handovers usually mean there’s only one “best” new AP to choose from, the rest are down in the noise floor. There can be some interesting/annoying overlap on intersections. The machine travels past an intersection and is suddenly exposed to a strong AP just around the corner, so it hops to that… and that AP almost instantly disappears as the machine passes beyond the intersection and it gets blocked by the tunnel walls. That really gives the client a workout and is just usually the result of poor AP placement.
Mine sites are constantly adding and moving APs around, most places have a few hundred of them in their production areas. Line of sight range is about 50-100 metres for each AP, and tunnels are never straight for much more than that distance so there ends up being a lot of hardware. Panel antennas to extend range muck up the handovers as the sudden drop in the signal as you pass into the backlobe often doesn’t give the client enough time to scan the channels and find a new AP. 802.11r would help but we deal with a huge mishmash of AP hardware across sites so we stick with omni antennas to keep it as simple as possible haha
This example is a real WTF. I really hope nobody is planning on building safety-critical real-time systems on top of WiFi!
I imagine many already exist. But the system should be designed to fail safe with WiFi in mind.
I don’t know about manufacturing environments but I deal with laboratories a lot, and I’m a bit baffled at how quickly lab operators have jumped on battery-operated wifi sensors for lab monitoring systems. I have like three room sensors attached to my EcoBee thermostat at home and I can barely be assed to change the batteries in those things, I cannot imagine dealing with batteries and connectivity troubleshooting for a building full of sensors whose reliable operation is often critical for regulatory compliance. Seems like the perfect application for PoE systems, to me
In industrial there are very few wireless systems unless they are either too remote from the CPU and aren’t safety sensitive. Safety is taken very seriously because any incident can mean injuries/death and ending up in the public eye. Any safety systems are hard wired because of reliability.
as a software developer, that example screams bad design
It screams “live service”
It also reflects something probably half the industry would push for since they can monetize it.
Are you new to the planet? Let me tell you about this thing we have called capitalism…
Better hope staff don’t Microwave their lunch at the wrong time….
If your robot moves around, then it needs a wireless connection. And it doesn’t really get any more reliable than wifi. I’m certainly not going to outsource that to a Verizon cellular connection.
And even for things that can be wired - ethernet is far from reliable. Cables are easily damaged or simply unplugged.
Wifi can work really well, especially with high end networking gear (and not, for example, the wifi access point you get for free from Verizon).
I think you’ve missed the point.
Anything automated that could be a threat needs to have safeguards. Needing constant wifi to prevent death or injury is not an acceptable safeguard.
Consider consumer/professional drones. If they lose connection they have on board protocols to mitigate hazards. Even then they are still governed by laws to isolate then from people because even those safeguards aren’t good enough. Suggesting that a robot could completely rely on wifi is preposterous.
I think the point is that that sort of safety critical stuff should be on board, not relying on a wireless connection.
Yesh it should be self contained. Although to be fair there shouldn’t be a way for a human to be there to begin with.
As someone using various wireless standards over over twenty years and in IT dealing with wifi instability on basically a daily basis. No.
Wifi is a series of compromises to be convenient. It’s “good enough” for most of those but generally and increasingly in newer standards, the compromise is to drop stability for things speed. You’ll see this to be the case in a lot of professional wifi gear that will transfer you to a lower standard if it sees weaker signals to improve stability.
To make that concrete, a problem with wifi in an office is an embarrassing “I’ll call back on my phone” but a factory floor that could be millions of dollars of downtime to restart an entire chain of machines. Hardened industrial wiring and connections is well established and wifi is just not at that level. The poorly formed example of the robot was trying to convey their intention to start addressing that level of hardening.
All that said, based on my experience reading ieee articles this is all exaggerated. in reality we’re probably just getting more stable video calls at higher bandwidths. Still a win for the help desk techs everywhere and people with a heavy wall making Netflix flaky.
This sounds like they’re talking about something specific. There was a guy that was picked up/crushed by a robot recently that is eerily similar.
As long as they have a delay counter which immediately shuts the robot down when it hasn’t been answered within a certain time period it shouldn’t pose much of a problem if it has an E stop. Just inconvenient when it keeps shutting down all day.
I work in autonomous vehicle engineering. That’s not even on the table for something we’d consider doing. But China is trying to enter the market hard, and I am less sure they wouldnt do that.
Lol my company runs safety critical systems for autonomous mining equipment over wifi and have been doing it for 15 years. You “just” need to engineer it correctly to take into account an imperfect link. Just like you need to to with ethernet, because every link is an imperfect link when you’re dealing with safety critical systems.
Mining might also be a better suit to the technology is general. No interference problems from neighbours :)
Do you get timeout/watchdogs trips all of the time or is it pretty reliable?
Generally it’s ok. Wifi handovers between APs are the main issue, our wifi clients have handover times of about 70 milliseconds or so normally which is fine with the engineering controls in place. A shift’s worth of operation usually involves a thousand or more handovers with no issues.
Our machines are run in exclusion zones, and every entrance to that area has a twin barrier setup that creates a “no man’s land” between machine and personnel. The width of that is dependent on the approach speed of the machine and the worst case scenario timeout of the safety system controlling that area. So if you stand on one side of the barrier and the machine comes at you full tilt and somehow loses wifi just as it hits its side of the barrier, it’s still guaranteed to stop before it hits you.
And yes, we rarely get interference, seeing as we usually operate several hundred metres underground. But you’d be surprised to see just how many wifi clients are floating around down there, most mining operations are heavily networked these days.
Wow, that’s really interesting, thankyou.
I was imagining one AP, not lots of APs as fixed infrastructure. Something about being underground and safety critical made me assume the place wouldn’t look like an office building.
70 millisecond handover is pretty good, I assume 802.11r? If you control the clients then all is good and fixable, but if they’re (thunderclap) external vendor controlled (lights flicker) I imagine some will keep trying to hang onto old out-of-range APs for too long.
The area we cover is usually up to a kilometre or two long.
We just use good old 802.11g. Our wifi clients are Moxa industrial units and I’ve found that they’re pretty consistently compatible with all sorts of AP hardware that sites use. Sites with Cisco hardware and controllers give us the most grief, Cisco loves Cisco and nobody else, of course.
As wifi is pretty much line of sight underground (it doesn’t penetrate rock), handovers usually mean there’s only one “best” new AP to choose from, the rest are down in the noise floor. There can be some interesting/annoying overlap on intersections. The machine travels past an intersection and is suddenly exposed to a strong AP just around the corner, so it hops to that… and that AP almost instantly disappears as the machine passes beyond the intersection and it gets blocked by the tunnel walls. That really gives the client a workout and is just usually the result of poor AP placement.
Mine sites are constantly adding and moving APs around, most places have a few hundred of them in their production areas. Line of sight range is about 50-100 metres for each AP, and tunnels are never straight for much more than that distance so there ends up being a lot of hardware. Panel antennas to extend range muck up the handovers as the sudden drop in the signal as you pass into the backlobe often doesn’t give the client enough time to scan the channels and find a new AP. 802.11r would help but we deal with a huge mishmash of AP hardware across sites so we stick with omni antennas to keep it as simple as possible haha
You could safely bet somebody already does