What's new
What's new

Feedback on concept: remote robot workforce

EAutomation

Plastic
Joined
Mar 12, 2018
Hey guys,

We're a robotics startup and we're looking to gather feedback on a concept: Remote Robot Workforce / "Robot Call Center".

"Robot Call Center" - Album on Imgur

Essentially we are making more affordable robot arms with remote human operators. The idea is that for a significantly reduced upfront cost (think $3K instead of $30K) and using remote human operators for a few dollars an hour, we can provide value to various manufacturers and factories. The robot arms themselves will have human arm length, a payload of a few kilos, sub-millimeter precision, and cameras to help guide operators.

Here's how we'd summarize the benefits:

- Low upfront cost: e.g. a $3K arm instead of a $30K+ arm. The particular arm we're looking to provide has specs similar to (but slightly less than) a UR5 (human arm length, 5kg payload, sub millimter precision).

- No need to program, easy to integrate. You can talk to the remote robot operator over Skype/similar, and explain the task to them, and assuming it's a fairly mundane and repeatable task, they should be able to start work quickly.

- Task adaptability. Related to the above point, a human can learn to modify a task with slight variations no problem (e.g. switch from packaging item A to packaging item B later in the day). Also reduces the need for a perfect controlled environment.

- Advantage of cheap labor. Many diligent workers around the world will be happy to operate a robot for your business for a few dollars an hour.

- Flexible and scalable. These are contractors - no need to keep them on after a contract ends, no need to pay benefits, ability to circumvent local labor regulations (to some degree).

How it works:

Our robot arm has a "neck" with cameras overlooking the robot's workspace, plus a wrist-mounted camera so the end-effector interaction is easy to visualize. The remote workers control the robot like a video game intuitively with their keyboard and mouse, using the robot's cameras to see what they're doing.

As an employer you post a contract job (with a description and an hourly rate), pick a contractor, then get talking to them directly over video chat.

Request for feedback:

Very happy to hear your thoughts on this concept, and your potential willingness to try out such a concept. If you could, please post a summary/description of where you work/have worked in a field with physical labor and production (e.g. metal shop, injection molding plant, chemical production lab, etc) and your work experience (e.g. manager, line worker, foreman, etc).

Quick background - our startup is going from prototype to production quality products soon, and looking to begin pilot testing our solutions in the right initial market(s).

Cheers =)
 
I am having a little trouble wrapping my head around this one. I suppose for a simple pick and place operation it might be worth while, but I cant think of many shops that would want several cameras connected to a global network poking around their space.

I work in aerospace, and dont think we would even be allowed to have a setup like that and still comply to customer requirements.
 
I am having a little trouble wrapping my head around this one. I suppose for a simple pick and place operation it might be worth while, but I cant think of many shops that would want several cameras connected to a global network poking around their space.

I work in aerospace, and dont think we would even be allowed to have a setup like that and still comply to customer requirements.

Thanks for your reply.

Yeah we understand there's going to be a wide range of use cases for which this would be prohibited. What we're really trying to do is pick off a few of the remaining ones where the value prop is clear and barriers are minimal, and then target those for initial pilot testing.
 
I think you are underestimating the required life safety issues.

Since these robots in essence are remote operated manipulators, all of the traditional controlled robot safeguards will be needed in addition to some other ones. These are the things that make robot installs expensive.

The concept of the cheap robot sounds good at face value but the 10X cost savings will quickly go away when you add in the wages of the operator added over time, no matter where they are located unless they are willing to work for free.

I don't really see any advantage for us Earthlings. From an owner standpoint, I am responsible for the operation of this automation. That means I need to have a significant level of confidence and certainty that whoever is on the control end is doing what they are supposed to do.

Having an operator that I have probably never seen or know does not make this idea very desirable. I would much rather spend the 10X cash for a known quantity with a specified performance level than an unknown operator and unspecified and unverified level of performance.

I think your idea is novel in that it would make jobs available for a large number of people with decent data link of some sort but I also think it will fail because cheap labor is often not cheap enough. You need to keep in mind why industry is so willing to utilize automation to they extent that it does. A huge issue has been eliminating the issue of human short comings. Cheap wages is not enough.
 
Thanks for your reply.

Yeah we understand there's going to be a wide range of use cases for which this would be prohibited. What we're really trying to do is pick off a few of the remaining ones where the value prop is clear and barriers are minimal, and then target those for initial pilot testing.

As any decent business would do. I don't see any value in the "fairly mundane task" segment. Any easily manipulated task can be carried out with simple mechanical/servo means, and it'll cost a small fraction of $3k. Heck, mundane tasks can be carried out with inclines and gravity.

What I'm interested in is seeing my chemical reactor parameters on my phone at 2:00am, my production machinery's parameters at 6:00pm before I sit down for dinner. I want the status of my manufacturing plant easily within reach remotely as I designed it to be autonomous in the truest sense of the word. THAT type of connectivity has more wide an application IMO.

I'm guessing EAutomation is a kind of engineering or business student start-up. Send me your coordinates via PM. I'd like to speak with you.

- Advantage of cheap labor. Many diligent workers around the world will be happy to operate a robot for your business for a few dollars an hour.

- Flexible and scalable. These are contractors - no need to keep them on after a contract ends, no need to pay benefits, ability to circumvent local labor regulations (to some degree).

The above quote however makes me apprehensive. In touting the above, you're doing your best to ensure Ontario remains a shit hole for businesses and employees alike.
 
Issues: Latency, confidentiality, delivery/install, and who picks up (whatever) when it gets dropped...
Even best-case latency of FaceTime calls, or the low-latency h.265 systems aren't quick enough to do things like back a truck up to a trailer hitch accurately. Huge-bandwidth systems, like the Cisco 'live' ones are going to be tough to deploy on the factory/shop floor where it's been documented thru countless posts that Spotify and basic 9600 baud serial transfers present shop-stopping problems.

Plus, customer-presentable robotics operators might cost more than a few dollars.
Plus again, wait 'til Trump finds out your illegal alien workers are border-hopping on "his" interwebs.
 
The advantage of a robot is not just a reduction in cost, but a reduction in errors, human errors. If you need to pick something up and move it exactly the same way thousands of times, a robot should be able to do that with 0 defects. If you introduce a human operator, a low skill, low wage, human operator, you're right back where you started.

You are trying to reduce the most complicated part of a robotic system, the automation. But, that's where the value is. The actual hardware matters very little. It doesn't matter that much if it costs $3k or $300K. If it can increase throughput and increase quality enough to justify its cost, the cost is irrelevant.

Taking a low wage, low value job and throwing a fancy remote control on it is not the answer we are looking for.
 
"Operators" for a few bucks an hour? What second/third world shithole are you going to source "operators" from? I see the day coming where "technology" displaces so much of the work force that the tired, poor,
huddled masses yearning to breathe free, rise in protest and blood runs in the street.
In the mean time, my favorite activity is slaying robots for the juicey goodies inside them.
 
+1 on the latency issue.

It's a clever idea, though. And sadly one that suggests that fewer and fewer of our kids will find it easy to find a job in the future. We already outsource some of our garbage to places like China. Solve the latency issue and what little hand sorting happens at recycling centers might go next.
 
"Operators" for a few bucks an hour? What second/third world shithole are you going to source "operators" from? I see the day coming where "technology" displaces so much of the work force that the tired, poor,
huddled masses yearning to breathe free, rise in protest and blood runs in the street.
In the mean time, my favorite activity is slaying robots for the juicey goodies inside them.

Like these ?

Last group is the "Scottish Delegation".....:D
 

Attachments

  • Minions-crowd.jpg
    Minions-crowd.jpg
    20 KB · Views: 59
  • Minions-1.jpg
    Minions-1.jpg
    2.7 KB · Views: 243
  • Scottish-minion_delegation.jpg
    Scottish-minion_delegation.jpg
    7.6 KB · Views: 243
Thank you everyone for your responses, I read and liked each of your posts. For context it'd be great to know more about your backgrounds so I know where the feedback is coming from. Justified dismissals of our concept is much appreciated, to help avoid barking up the wrong trees.

Some interesting points of discussion -

The concept of the cheap robot sounds good at face value but the 10X cost savings will quickly go away when you add in the wages of the operator added over time, no matter where they are located unless they are willing to work for free.

From our research we found that the setup and integration costs of a robot are anywhere from 3x to 10x the cost of the hardware. So let's say you want to set up a half dozen UR5 robots to automate some part of production. Well, by the time you're up and running that can be something like that 3*(30k/robot * 6robots) = $540K, i.e. an investment on the order of a million dollars (don't forget add-ons, maintenance, tooling, etc).

One of the big themes we work on is changing this equation for smaller businesses that wouldn't traditionally use robots.

Having an operator that I have probably never seen or know does not make this idea very desirable. I would much rather spend the 10X cash for a known quantity with a specified performance level than an unknown operator and unspecified and unverified level of performance.

I think your idea is novel in that it would make jobs available for a large number of people with decent data link of some sort but I also think it will fail because cheap labor is often not cheap enough. You need to keep in mind why industry is so willing to utilize automation to they extent that it does. A huge issue has been eliminating the issue of human short comings. Cheap wages is not enough.

Fair enough, can definitely see these issues being true for many industries. BTW while I didn't explicitly mention it, the robot does function just like a typical easy-to-program collaborative robot (such as the UR5) but without the fancy safety features and a slightly reduced set of specs.

So the remote operation is not necessary! :P But it's something we're gathering feedback on.

As any decent business would do. I don't see any value in the "fairly mundane task" segment. Any easily manipulated task can be carried out with simple mechanical/servo means, and it'll cost a small fraction of $3k. Heck, mundane tasks can be carried out with inclines and gravity.

Maybe for someone mechanically inclined such as yourself :) It's been 60 years since the invention of robots and we still haven't solved burger flipping economically without humans, haha.

What I'm interested in is seeing my chemical reactor parameters on my phone at 2:00am, my production machinery's parameters at 6:00pm before I sit down for dinner. I want the status of my manufacturing plant easily within reach remotely as I designed it to be autonomous in the truest sense of the word. THAT type of connectivity has more wide an application IMO.

We're not focused specifically on internet-of-things integration of other machines. But what we can do is give you a remote login to any robot operating into your shop. Get a first-person view of what it's doing through its cameras right on your phone/laptop. This feature would be available remote operator or not.

The above quote however makes me apprehensive. In touting the above, you're doing your best to ensure Ontario remains a shit hole for businesses and employees alike.

Whether it's immigrants, remote workers, or AI, the world is going to continue getting more competitive and there's nothing to be done about that. If it's not my company contributing to this, it's going to be another one.

Also it may be worth considering various "progressive" policies like the minimum wage hike being imposed by our liberal provincial government. $15/hour is only going to lead to more of the following:
mcdonalds_automation.jpg

Issues: Latency, confidentiality, delivery/install, and who picks up (whatever) when it gets dropped...
Even best-case latency of FaceTime calls, or the low-latency h.265 systems aren't quick enough to do things like back a truck up to a trailer hitch accurately. Huge-bandwidth systems, like the Cisco 'live' ones are going to be tough to deploy on the factory/shop floor where it's been documented thru countless posts that Spotify and basic 9600 baud serial transfers present shop-stopping problems.

Plus, customer-presentable robotics operators might cost more than a few dollars.
Plus again, wait 'til Trump finds out your illegal alien workers are border-hopping on "his" interwebs.

Yeah there are certainly technical issues that will make continuous real-time operation challenging. What we might do is somewhat of a hybrid where an automation path is created and repeated in-house by the user, and then a virtual remote assistant can check in and "babysit" the robots, taking manual control if/when something goes off-course.

And if that kind of attention gets drawn to what we're doing... well that's almost a good problem to have :P

The advantage of a robot is not just a reduction in cost, but a reduction in errors, human errors. If you need to pick something up and move it exactly the same way thousands of times, a robot should be able to do that with 0 defects. If you introduce a human operator, a low skill, low wage, human operator, you're right back where you started.

You are trying to reduce the most complicated part of a robotic system, the automation. But, that's where the value is. The actual hardware matters very little. It doesn't matter that much if it costs $3k or $300K. If it can increase throughput and increase quality enough to justify its cost, the cost is irrelevant.

Taking a low wage, low value job and throwing a fancy remote control on it is not the answer we are looking for.

I appreciate your point, so I should probably clarify that we're looking to target smaller businesses and industries that may not even be thinking about using robotics in the first place. For many of them, yes, $300K matters a lot as an upfront cost, especially if this robot asset needs to be re-programmed (which is not a trivial cost either) every few months when a product line changes. We're probably targeting a greater proportion of the 90% of North American businesses that are 20 people or less, relative to traditional robotics companies.

To come back to this somewhat lame but illustrative example, consider burger flipping. You don't need hardware with a 0.01mm repeatability, nor the complex programming typically associated with said systems. But you also don't want to spend $300K upfront. And there's probably room for a simple arm with remote operation to save money over increasingly high minimum wages. These are the sort of opportunities we're looking to learn about and explore.

"Operators" for a few bucks an hour? What second/third world shithole are you going to source "operators" from? I see the day coming where "technology" displaces so much of the work force that the tired, poor,
huddled masses yearning to breathe free, rise in protest and blood runs in the street.
In the mean time, my favorite activity is slaying robots for the juicey goodies inside them.

:crazy:

Quick example: the average Romanian in manufacturing gets paid about $400/month. Dividing by 40 hours/week * 4 weeks/month, that comes out to $2.50/hour.

We understand there are A LOT of limitations with remote work (language, quality, experience, cultural expectations, etc). But there's a lot of potential for particular jobs to arbitrage value and opportunity. For example websites like Upwork.com have allowed us to get exceptionally high quality graphic design work and marketing materials for a tiny fraction of what the fancy folks in San Francisco wanted to charge - and frankly the quality was more than good enough.

+1 on the latency issue.

It's a clever idea, though. And sadly one that suggests that fewer and fewer of our kids will find it easy to find a job in the future. We already outsource some of our garbage to places like China. Solve the latency issue and what little hand sorting happens at recycling centers might go next.

Latency is certainly an issue. On the technical/product side, we'll probably look into segmenting parts of remote operation into "quality assurance" and "manual control when necessary".
 
OP wrote: "What we might do is somewhat of a hybrid where an automation path is created and repeated in-house by the user, and then a virtual remote assistant can check in and "babysit" the robots, taking manual control if/when something goes off-course."

Aside from scaling the process, why would I need an outsider to babysit and take manual remote control when I, the user, created the automation path in-house? When I walk by it and see all the parts on the floor instead of the machine, I can see that something has gone off-course. Proper path design utilizing a 'task complete' action/marker/switch/flag/whatever will let a local system know things are good or bad just as easily as letting a remote system know. Maybe producing low-cost, easily-programmable robots with good integration tie-in to legacy local area systems is the way to go. Crowd-source your delivery, install, and initial setup/training with customer-presentable professionals, who can monitor as needed, visit on-site if need be, and always have a few backup units in the truck. There's a business model I could get behind, or even participate in for a few years. Having to re-explain what's wrong to successive layers of Hungarian tech support (no offense) is not a great way to spend your afternoon.

In my opinion, of course...

Just got a newish Volvo wagon for my wife. Learned that Volvo's business model is shifting from 'hey, buy this car' to 'hey, buy a vehicle subscription (for approx. the same price) and we'll have the car you want up and running all the time. If it dies, we'll bring you another and cart away 'yours' for either repair or replacement. In the future, if you need a 9-passenger SUV instead of your little convertible for a month, we've got you covered. Just book it, and we'll bring it to you and debit your account if there's a cost difference.'

Apply to small-scale business robot users. What you're selling is process uptime, not machines. Big shops will already have all their own guys and expertise, and won't really need what you're selling, so a hugely-scalable model per-customer isn't needed. You need hugely-scalable per-market, but for you.
 
We understand there are A LOT of limitations with remote work (language, quality, experience, cultural expectations, etc). But there's a lot of potential for particular jobs to arbitrage value and opportunity. For example websites like Upwork.com have allowed us to get exceptionally high quality graphic design work and marketing materials for a tiny fraction of what the fancy folks in San Francisco wanted to charge - and frankly the quality was more than good enough.

First, if you can basically undercut the UA5 for $3k, that's a pretty solid business in and of itself. Perhaps why even bother with the human aspect? Ditch the custom pendant hardware for an iPhone/Android. Let me remote to the arm with it to check on my process, that would be neat.

Second, you can get around many of the issues of culture/language by using US based workers. We've got a few million people on SSDI who are on the doll because of physical injuries that prevent them from doing manual labor, in locations where office/sit-on-your-ass jobs are hard to come by. Get the lawyers to do mass independent contracts, or work with a state so you can bring the wages down to something less than minimum wage in an Uber-Like scenario where these folks can work from home. Yes, you'll pay more than you will for $2.50 an hour Romanian labor, but the cost savings from language compatibility, cultural overhead, bandwidth, etc will more than make up for it. You could even background check some of them so they could remote into the aerospace type shops... and people will feel WAY more comfortable if they know it's Frank in Tennessee who's overseeing their robots instead of Oleg in Transylvania.
 
Hate to say that but I hate robots. A man who goes to replace a man by a device is a disgrace to the human race.
 
Hate to say that but I hate robots. A man who goes to replace a man by a device is a disgrace to the human race.

It's not even the robots that is the problem, it's the way their going about it. Given the total lack of understanding what a responsible business is, and the emphasis on competition and competitiveness, they're likely a lot of engineering students. The photo of McD's says it all- application may be retail, or sorting and packing, but nothing to do with manufacturing given the simplistic design constraints. Green software and electrical engineers often think that automation is the key to all of businesses' problems- the technology has been around for 60 years to do what they're proposing (in rudimentary form, granted, but has existed nonetheless). If it had any promise, it'd be implemented (and is, just not in the form factor they're pitching).

My request for contact was met with a PM, "What's up?". No company name, information, profiles, not even signed by anyone. Maybe their robots are writing their emails?:scratchchin: Good way to approach a customer. Preliminary searching reveals the company as written in the PM handle doesn't exist.

The only info I got was they're in "Toronto". Which could mean an area of hundreds of square kilometres from King City to Milton to Pickering, ON. I wanted more information on their approach- but I'm not going to beat a path to their door (Hint: neither will anyone else).
 








 
Back
Top