This is a very big deal. Amazon has been looking for this technology for years. They've held competitions for it. That video is one of the best unstructured bin-picking demos ever seen. The combination of a four-fingered semi-rigid gripper and a suction cup seems to work well. It's fast; many vision-guided systems are still painfully slow. Amazon should be able to pick everything in their inventory that weighs less than a kilogram, is reasonably rigid, and has a surface that can be grabbed with vacuum. Everything lightweight which comes in a box, can, or jar qualifies. That's probably at least 75% of the inventory.
This should destroy about 50,000 to 100,000 jobs a year for the next few years. That's more than the entire coal mining industry.
Yeah. I worked next to these guys in the Harvard iLab when they were getting going. They had the opportunity to cut corners and cash out early on, but seemed to remain focused on building a truly robust, broadly capable picker.
I often noted with them at the time similar notions of the implications for the job market. There's so much hype out there, but I think RHR will be one of the companies that actually illustrates to the public where things are headed.
This type of picker could be used in so many more applications outside of warehouses. Help disabled shoppers, pick fruit, litter picking, sorting recycling, etc etc
Don't think that first one is going to happen any time soon, but definitely the rest are all probable.
Having motorized components near people sounds like a recipe for disaster. Especially around people who might not be fast enough to move out of the way if something goes wrong.
It will mean loss of jobs (however low paying and not fulfilling it may be) for people who can least afford to. People that cannot hope to retrain for a better job or have some kind of retirement plan in place.
I saw an orange in that bin. I wonder if they are trying for agricultural applications. Fruit picking is very labor intensive...and some fruit and veggie picking remains difficult to automate, especially the more delicate items.
Put into perspective: A picker in a warehouse costs conservatively something like $25/h (labour costs, management, HR, breaks, hiring costs, etc).
$25/h * 2000h/year = $50,000/year. Any robot that can replace a single picker that costs less than $250,000 would earn back it's capital in five years. And this robot? I don't think it costs a quarter million. I bet it costs in 5-figures.
The biggest barrier will be if they can produce these fast enough to meet demand.
A picker only works 8 hours a day. That robot will work 24/7 till it breaks. Which means it will pay for itself in one and a half years, even if it does costs $250K.
I think you're wrong about $25/hr though, it's probably closer to the minimum wage. So maybe 3-5 years is the right estimate after all.
Back in my MechE days, I'd buy everything from McMaster-Carr. Probably overkill for little desktop robots but if you want to build things that are the size/power of an industrial arm it should have what you need.
In the short run, (next half a century) this will be a disaster for the least privileged. In the long run though it will mean liberation from demeaning hard labor in the same way that tractors and industrialization did the same for our ancestors.
17 comments
[ 3.9 ms ] story [ 59.5 ms ] threadThis should destroy about 50,000 to 100,000 jobs a year for the next few years. That's more than the entire coal mining industry.
I often noted with them at the time similar notions of the implications for the job market. There's so much hype out there, but I think RHR will be one of the companies that actually illustrates to the public where things are headed.
Having motorized components near people sounds like a recipe for disaster. Especially around people who might not be fast enough to move out of the way if something goes wrong.
I have mixed feelings about this...
[1] https://www.youtube.com/watch?v=mS0coCmXiYU
$25/h * 2000h/year = $50,000/year. Any robot that can replace a single picker that costs less than $250,000 would earn back it's capital in five years. And this robot? I don't think it costs a quarter million. I bet it costs in 5-figures.
The biggest barrier will be if they can produce these fast enough to meet demand.
I think you're wrong about $25/hr though, it's probably closer to the minimum wage. So maybe 3-5 years is the right estimate after all.
https://www.iamrobotics.com/
They seem to rely on a stereo vision that perhaps is a CMU spin-off?
I hope I'm proven wrong.