Boston Dynamics’ new robot stacks boxes [video]
youtube.com
youtube.com
The roboticization of manufacturing is mostly finished since factories work with identical outpus SKUs every time. Warehouses tend to have to handle a lot of different SKU but in a fairly regular way meaning that they're an environment that's pretty much ripe for robots to be entering just now. Goods to picker systems (like Kiva) have really taken off allowing robotic pickers like the one sold by us a nice ecological niche to fill. There are also companies with robotic forklifts and all sorts of other things.
I'm not entirely certain that the role these bots are performing wouldn't be better served by a large stationary robotic arm, as some other companies are working on now. It might very well be the best solution for unloading trucks?
This was my reaction, though I suppose it's a capability-demonstration video. Would think in unloading, you'd want something heavier to unload full pallets/carts etc.
Possibly, they'd be useful for rare-item box picking in DC's, where the runs are long enough and infrequent enough that it's costly to carry a full shelf (slower transit speed, plus need for return trip)? Or maybe it's just a capabilities demonstration and the market caught on to the whole search and rescue thing [0].
Optimization is about what can you assume is "fixed" about the environment (and so not engineer a solution around it) and what must you consider "variable". Clearly fixed constraints are easier to engineer around, and if all of your pallets and racks were arranged similarly, then a robot you could move somewhere, have it extend its "base" to provide stability, and then run all day would be a win.
The other variable then is how often do I do this task from this space. Assuming setup and tear down of the robot takes some number of minutes/hours, those add as a cost to change tasks. If the robot is mobile and can move itself from task to task, that changes its economics as well.
So lets assume as an example, that there are six bays (loading docks) at the warehouse and each bay of the warehouse receives one truck a day which offloads up to 6 pallets that need to be transported on to the conveyor.
If all trucks arrive at the same time:
Six robots will efficiently unload them as soon as they arrive. (optimizing for truck parking time) At the expense of six robots.
One robot that can be packed up, moved, and unpacked, can work all bays at the cost of loading bays that cannot be reused until all previous bays have been unloaded. How long does that take? And how does it affect how many goods can be moved through the warehouse? (which is how money comes in)
Then there are trucks that arrive at different times. Same options but now you have idle robots (if you have six robots) which are waiting for a truck. Although you might be able to overlap robot movement with truck arrival to keep your bays reasonably busy.
At the outermost level of the problem (where the business people look at it), a warehouse is essentially a "switch" that moves goods from bulk carriers to more specific carriers. Each good moved through earns its percentage of the transport cost. Each good incurs a storage cost while sitting in the warehouse. Lots of metrics around how many times you can completely fill and empty the warehouse of goods vs the money you can get for transporting goods.
Bottom line is that reality (and logistics in particular) are pretty complicated and I expect there are times when a stationary arm would make more sense than a moving arm. However I consider the 'moving arm' to be the super class of 'arm' (to me it seems easier to turn an arm that can move on its own to one that is stationary, and much harder to give a stationary arm the ability to move.)
The arm itself may be great. If you need it to be mobile, then there are much lower-tech solutions to that part of the puzzle, like a modified forklift. You can still have a computer drive it around if that makes sense.
Do not step inside this perimeter or you may die.
If a 200kg robot arm gives you a love tap because you were just barely in the perimeter you may not survive.
We’ll fix it in post is ok for film production. Move fast and break things is ok in free to consumer services or SaaS where 0.99 uptime is great. It is not ok with industrial automation.
Also, how well does your sample match the US factory population? Is there data on robotization of factories by size, product type, etc?
To me, saying robotization is complete is akin to saying computerization of business is complete because IBM has a mainframe.
I did a tour of the Cabot cheese factory in Vermont, and they have a fixed arm palletizing cheese wheel cartons. It looked a lot more reliable than these robots and seemed to have fewer moving parts.
Pickle Robot is also working on this but with a fixed base and support for random box sizes/weights. We’ve just started scratching the surface of palletizing/loading with reinforcement learning.
If you are excited to come solve robotic box handling ping me! aj at picklerobot.com
Boston Dynamics is developing amazing tech but they are demoing robots for tasks that are better with 4 wheels. Dynamically balanced walkers inside are niche application.
Smarter robot hands and 4-wheeled robot movers are where the markets and money is. Boston dynamics may be moving towards that direction but they still like to show the cools stuff they can't find markets for.
DARPA funding is not necessarily violent. One can argue that the current explosion in self-driving cars is due to the DARPA challenges in the 2000s.
I had the same intuition, but then I remembered the toy I'm playing with - the spectacularly unstable 4-rotor copter that could not even take off without an onboard computer actively keeping it level all the freaking time. And it's rock solid almost no matter what.
As technology changes, our intuitions must follow.
2-wheeler is not better or more stable than 4-wheeler and requires more maintenance.
if the 4-wheeled version must also have the same level of acceleration and speed of picking up, won't it also need a counterbalancing mechanism? It just won't need the "balancing algorithm", but will need a way to balance the arm swinging out.
Also, the counterbalancing mechanism is written with software, so once perfected, it's "free" to deploy. The weight could be the actual batteries, so no added extra weight!
Do you see cars with huge counterweights?
And no, it is not free to deploy by any means. Any moving, huge, swinging parts will cause expensive problems. Remember your hard disks.
Of course it depends on the weight of what you are moving, but it doesn't seem like the 2 wheeled robots in the video can actually move "heavy" weights.
I would not call 15 Kg "heavy". I can carry that myself on a single arm...
Wheel issues (typical):
1) flat tire (if the wheel is not "solid" only)
2) axle bearing
Controller+swinging counterweight:
1) a gazillion electronic components, connectors and contacts, bearings as well, the actuator for the swinging, etc., etc.
What happens if the environment is a bit humid?
For the wheel, a little rust, for the complex electronic system?
And BTW, we can save 25% by using only three wheels ...
For a device leaning into shelves to grab things, being dynamically balanced might be useful. Like a racing sailboat with a movable keel or ballast. But you could do that with a three wheeled design.
I bet there's also a static version of this bot with three wheels, slightly larger footprint, where the battery pack is situated over the third wheel for leverage.
But the swinging battery pack? In theory it reduces the footprint of the bot, but any time the bot is trying to grab a package, the battery is moving. Which means you can't let another robot or human enter that space. So either you have a very very complicated calculation of what areas of the warehouse are 'empty' at any moment, or your logistics program has to treat the robot like it's its worst-case footprint at all times. At which point you might as well make it static. Right?
Seems like an easily enough problem. Basically every car out there has reverse sensors...
This is a smart usage of a dead weight (batteries) that makes all the other motors smaller and lighter.
Thanks for teaching me something :)
Source - I was part of a project in college that built a recumbent tandem bike. We built the frame out of large-diameter steel tubing, and it was quite heavy. I followed it on a normal bicycle several times while testing. Even with 2 people pedaling, it was easy to keep up with going up hills. It would tend to get away from me going downhill though.
I think they have solved 2-wheel control. Lets move on.
This one for example has an organic birdlike head motion. Is that pre-programmed or a natural outcome of their control system dynamics. The MIT technologyreview page says "Each one is created with carefully pre-programmed movements and will take many, many takes to get right before it’s shared."
So is it a sequence of individual actions: align to pallet using registration cards -> move forward with reaching distance -> detect box position -> use box picking algorithm to pick box -> backup fixed precalculated distance -> turn 90 degrees -> move forward to destination etc. This would be a brittle solution that has to be reprogrammed for the next task.
Or is it a more general closed loop solution: Theres a pile of boxes there. Move it here. Use your sensors to plan a trajectory and sequence of actions to accomplish the goal.
That's what I want to know.
Similar problems are likely to elicit similar solutions.
Also, I'm watching some lectures for a MIT underactuated robotics class (https://www.youtube.com/watch?v=VeEqtTgDXFc). The professor says in one of the videos that the people who end up creating Boston Dynamics also took that course. From what I watched until now, nothing is explicitly programmed (by some vague definition of "explicitly programmed"), you just find some inputs that makes the dynamics of the model behave the way you want.
Almost certainly the birdlike movement is the outcome of the robot dynamics, and not some hardcoded rules. But things like avoiding hitting obstacles or estimating the distance, I think that needs to be explicitly programmed (and then used as input to the model such as "move the end effector to that position").
Yes, it makes sense to mimic nature sometimes. But when it comes to robotic manipulators, loaders, etc - there's absolutely no reason to copy biological systems limited by strength, speed, or joint mobility. They seem to be deliberately going for that uncanny valley effect.
Eventually, someone is going to admit it. Until then, this is hugely entertaining.
Perturbations do exist and they add up in the PID feedback loop. For example you can see in the video that the way each box is handled varies and that the robot reorients itself to pick/drop boxes.
Consider a simpler 1 dimensional example: https://youtu.be/JpNAhKT7yY4?t=101 (warning, loud). You can observe the system adjusting to the hits by the person, but also when left alone it will have to adjust to accumulating externalities like gravity. Also note the natural/biological looking swaying motion that results.
You know, sometimes you see a solution, and it's just so elegant... This isn't elegant, or brilliant. This is an attempt to solve self-imposed, artificial problems.
P.S. people who liked your pendulum video (I did), might also enjoy this drone pole acrobatics video:
Also: Kept waiting for some guy to come in and try to kick it over. I thought that was Boston Dynamics' signature move in these promo videos...
When the robot revolution comes, those guys at Boston Dynamics are definitely going to be the first up against the wall, without doubt.
Robots rights or something.
Static doesn't necessarily mean safer. But the article also notes these aren't intended to share space with humans yet.
Also, recall from freshman physics that the pendulum problem is non-linear. Most balancing robots linearize the problem around the stable point. Which means if the robot gets far away from the equilibrium point, you either have to switch to a different controller or live with the consequences. One of the consequences is the robot racing off to the horizon as it tries to get the wheels under the CG again. (Which may happen anyway whether or not you switch controllers.)
Thus, my standard advice to all newbie robot builders: Your first robot should have a low enough mass that a missing semicolon will not punch a hole in your living room wall.
If this kind of thing is interesting to you, MIT 6.832 Underactuated Robotics by Russ Tedrake is posted on YouTube going a few years back, and the Spring 2019 version is being live-streamed every Tuesday and Thursday, picking up again next Tuesday after spring break.
* If the robots need to "reach" for items, you have to figure out balance/stability issues regardless of how many wheels you have.
* The fewer the wheels used, the smaller the footprint of the robot.
4 wheels can be inherently statically stable but requires more space and requires more energy when running (counter-intuitive, but if your balance algorithm is efficient, you use less energy wheeling around on 2 than on 4 because you can basically play games with gravity to get some acceleration tradeoffs... And you have half the number of powered wheels, assuming you're 4-wheel drive).
If they solve the hard math, the end result may very well be a more performant machine (though as someone else in thread has noted, depending on the overall goals, a zero-wheels statically-mounted arm could be better still).
Whatever BD's angle is, it's not efficiency.
When the engineer is approached he tells the audience that the assistant just switched the robot back to manual for safety reasons.
I think no one could have noticed and I like that kind of honesty. I also found it telling that the people who built the robot have respect for its dangers.
They will just have to remove the people to make it more efficient.
https://www.youtube.com/watch?v=EeMTZd68fOU
The robots were above the boxes instead of below the shelves (maybe to beat an Amazon patent? I dunno.) But that warehouse looks like nightmare fuel for a maintenance crew.
> The blaze began high in the warehouse and was so difficult to reach that firefighters had to cut holes in the roof, according to Neil Odin, chief officer at Hampshire Fire and Rescue Service. A sprinkler system only contained the fire in its early stages. It took more than 200 people to quell the flames.
https://www.uline.com/CustomerService/ULINE_FAQ_Ans?FAQ_ID=4...
Completely retooling the society after 70 years of postwar infrastructure is going to take more than anyone would willingly sacrifice. You would have to bomb everything more than a few years old flat. At no point will it be feasible to keep leaving major investment in business and industry tied to the promises of modernizing through bleeding edge technology. The world is so vast and involves such a colossal distribution of existing resources. So many of the systems here are only functioning through broad and cheap standards, like containers, rail gauges, ship sizes.
I don't want my job. I want a machine built to put me out of it for good. I have been saying that along with a generation of workers told robots were coming for their jobs since the day i was hired. That was fifteen years ago.
We need a Manhattan project for the basic tools of industry. Now. There are thousands of novel or unproven methods of doing the most basic forms of labor our economy is based on, and we are leaving the creation of things that need to be ISO standard across the surface of the earth in order to succeed to entrepreneurs and startups and scholastic vanity.
The modern technological landscape across all disciplines looks terrifyingly similar to the cambrian explosion, which produced so many things, at such a cost of living suffering, that did not survive what came after. I'm afraid our civilization might have run out of low-hanging fruit.
To me that is the line of demarcation between the "developed world" and whatever precedes it. I don't think what is beyond that is peaceful. It necessarily undermines the infrastructure the whole society is founded on.
Look, we're rapidly approaching a point where all human needs (food, shelter, etc) can be handled by automation. If we reach that point, why do people need to have jobs? Why can't we turn our lives towards something better?
Because we will likely still be living in capitalist societies in which people increasingly unable to provide value to the marketplace due to automation will still need income to survive, and no one implementing that automation has any intention of doing so in order to usher in some kind of post-scarcity collectivist utopia.
Automation exists to allow the ownership class to reap the benefits of labor without the burden of compensating human employees, not to free the labor classes from having to work.
I’m also a writer and I study the intersection of automation and society.
So while it’s true that most automation engineers don’t think this way, some of us do. :-)
The problem with seeing automation as a means to liberate people from capitalism and free them for intellectual pursuits (a noble and laudable goal) is that it exists to serve capitalist ends. Amazon, for instance, is definitely not automating for the benefit of humanity. They just don't want to pay people.
Although it is good to know someone is thinking about it. I really worry that, at scale, mass employment won't be addressed until it actually threatens the status quo.
But the technology, while not liberating, will become more accessible. And I think a lot of people, just as in many decades past, are interested in something beyond bare capitalism.
My hope is that the collectivists can make progress in the coming decades too. We can develop our own open source alternatives to the machinery of dystopia. We can abandon intellectual property as a concept and ensure that when we labor for new development, the knowledge we gain is accessible to everyone who wants to liberate themselves.
And just as a group of hobbyists and corporations built Linux, perhaps we could build our own alternative to dystopia.
Politically right now I think we need unity and we need to stop trying to use the government to solve our problems. The government is a tool of force and force cannot be the way we build utopia except perhaps as defense against those who would have it shut down.
Andrew Yang. See his discussion with students at Georgetown Univ. https://www.youtube.com/watch?v=8tuJ0phjFys
or for a longer exposition of his ideas see https://www.youtube.com/watch?v=cTsEzmFamZ8
Another question: what proportion of homes had washing machines by year starting in 1928? Large proportions of American homes weren’t even electrified in 1928.
So I don’t believe an automated utopia is going to happen automatically. We’ve got to work to create corporate structures that naturally reward all of the workers instead of a few at the top. Cooperatively run companies are one way to do this. If the major companies of our world were cooperatives, a wide swath of the population would gain the wealth our world is increasingly producing.
But technology alone won’t solve the problem. We need to intentionally structure our society in a way that makes this socialist utopia possible. That’s what I study and I’m trying to figure it out. I’m also a robotics engineer.
I'm interested in reading more about this. Do you have books/research papers you recommend?
I’ve not read Karl Marx but I’ve learned from others that his critique of capitalism is worthwhile and relevant today and has been since it was written.
At one time I found Richard Wolff helpful, including his book “Democracy at Work” about worker run collectives. His lectures may be of value but lately I find him hard to listen to because of how kind of smarmy he gets.
This interview with Murray Bookchin from 1986 was pretty informative: https://youtu.be/kBP_BMOblzc
Otherwise the subject of interest is called “anarchism” or “libertarian socialism” or “communism”. It all falls under the broader umbrella of “socialism”, which is the intentional organization of society to produce a certain quality of life for all people. All of those terms have been heavily loaded by history and it’s propaganda, so you have to search carefully. Pundits of all persuasions will make broad claims about these social sciences, but know this: socialism and the other ideas laid out have a rich and complex history that absolutely must be researched. Ignore anyone who says the soviets were socialist and they were horrible so socialism is horrible. There is a history of socialist movements going back over 150 years that includes many writers and activists including those who were murdered by the soviets or the nazis precisely because they wanted power in the hands of the people, not the state.
When you look this up, also study imperialism, feminism, the black panthers and black liberation moments, lgbtq movements, and liberation movements in other countries. If you’re white and/or otherwise privileged (as I am) know that the experience of less privileged people are so different from yours that they absolutely cannot be excluded from the development of a utopia.
For more information see here: https://theanarchistlibrary.org/
There's going to be a tipping point where the distribution of wealth is so unbelievably bad that voters will finally start voting for policies that will redistribute more equitably.
Moreover: liberating technology will not be developed by the capitalist establishment because our liberation runs counter to the opportunity we represent as tools to be exploited. This is why Silicon Valley relies on surveillance and the big companies don’t share their key source code. They don’t have the goal of liberating us.
Invention of technology doesn't rely on capitalist establishment though, as long as there is at least one engineer out there to work on it just for the sake of it. For example, the invention of linux.
The reason this has happened is that we use an economic system where laborers are paid market rate and the owners are paid the rest. So when we automate something new, the market rate for labor stays the same (at best), but reduced costs mean there's a bigger chunk leftover for the owners. So business owners capture all the returns from automation. As long as this holds, you should expect increased automation to benefit whoever already owns the machines, not society at large.
And there are lots of possible solutions for this, some quite simple, but they're all considered socialism or communism, so when you start talking about them people assume you want Soviet Russia and stop listening.
Bucky Fuller wrote "GRUNCH of Giants" which says something similar.
("GRUNCH" is GRoss UNiversal Cash Heist)
https://fullerfuturefest.com/humanitys-critical-path-from-we...
How do you imagine we manufacture for example microprocessors in such scenario? I.e. how does a group of people start the next Intel as a co-op, without ever taking any external capital? BTW I've sent the same question to Chomsky some time ago, but no response yet.
Either a new industry, or new innovations in existing industries, creating more low-skill jobs in the same manner that previous industries have done.
Or make the need to work in order to make a living obsolete for many.
Maybe some people actually like that stress and pressure to succeed (lest they starve); takes all kinds.
>why do people need to have jobs? Why can't we turn our lives towards something better?
Answer: the basic faults inherent in basic human nature that have prevented us from getting along at any point in human history : avarice(greed), lust, sloth, envy, gluttony, etc. General religious intolerance and ethnic hatred, war, prejudice, etc. And that's just among the so-called normal people. Factor in psychopathic behavior and it's not difficult to find an answer. But that only covers behavioral issues.
Now factor in contributions from natural disaster, drought, pestilence, pollution, scarcity, etc. and this explains why I don't think we have a chance.
Because humans want more than just basic human needs addressed?
Everything that died in the Cambrian explosion would have died anyway if they had all been the same species; but having that variety available allowed the best designs to be selected. We wouldn't have humans without it.
I am still baffled as to why Cold Storage today is still not fully automated.
it's great if you're hungover.
> We began as a spin-off from the Massachusetts Institute of Technology, where we developed the first robots that ran and maneuvered like animals. Now we are taking the next step, combining the principles of dynamic control and balance with sophisticated mechanical designs, cutting-edge electronics, and software for perception, navigation, and intelligence.
They seem pretty clearly to be a R&D company around dynamic control and balance, so it makes sense that they would have rather complicate robots.
But once in a while I see a company run by engineers, and I just shake my head and back away. BD seems like a "tech utopia", with all the bad connotations that can have.
I understand energy efficiency is an odd thing to optimize for at this stage, but that would explain the design.
Handle is a robot that combines the rough-terrain capability of legs with the efficiency of wheels. It uses many of the same principles for dynamics, balance, and mobile manipulation found in the quadruped and biped robots we build, but with only 10 actuated joints, it is significantly less complex. Wheels are fast and efficient on flat surfaces while legs can go almost anywhere: by combining wheels and legs, Handle has the best of both worlds.
Handle can pick up heavy loads while occupying a small footprint, allowing it to maneuver in tight spaces. All of Handle’s joints are coordinated to deliver high-performance mobile manipulation.
https://naturalishistoria.files.wordpress.com/2013/03/terror...
https://en.wikipedia.org/wiki/Phorusrhacidae
A robot like this, able to move on 2 wheels, could be quite versatile for combat in tight spaces. Giving it a 4 wheel bogey like on an iBot, with the ability for the bogey to fully rotate, would also enable stair climbing.
https://msu.edu/~luckie/segway/iBOT/iBOT.html
The single manipulator could be mounted with a gun/bayonet, which could also be turned around to provide a surface for a door breaching ram. The thought of such a thing is pretty terrifying.
put an ak47 in that robot's manipulators and install some level 3 armor on the important bits and you've basically build a automated mobile weapon of terror and oppression
i don't remember the source but the only reason big dog was rejected by the us military was the sound it generated + power requirements. not really a problem if it's also firing guns at the same time and is deployed from a mobile self driving charging base. maybe the robot charges itself from the nearest tesla power station...
Palletizing is a routine robot task.[1][2] This is just palletizing with a mobile base. Now Boston Dynamics has to compete on price with the twenty or so other companies that do palletizing. If this turns out to be cost effective, Kuka will probably do it, too.
[1] https://www.youtube.com/watch?v=DiuFkMkReSs [2] https://www.youtube.com/watch?v=EgiUWycNFC0
But of course the part makers will say they are super secure and will have a PR making them look like as legit as google or amazon today. Because it's convenient, people will do it, considering the warnings comming from tin foil conspirationists.
Soon society will expect the level of productivity those parts give you, and most public and private services will assume this interface so the people not doing it will be more and more excluded from the regular system.
Isn't that happening for decades now ?
https://www.youtube.com/watch?v=eJ3RzGoQC4s
> the people not doing it will be more and more excluded from the regular system.
Same with cars, phones, internet
Why have the chosen to build it with only two wheels? It's clearly unstable and has to constantly adjust it's position and counterweight to stay balanced. Sure, it's super clever and I think it looks amazing. But surely it would have been far cheaper and easier to build it with three of four wheels so it naturally stable. It would have looked less cool, but I feel like that can't be the only reason.
Literally the only thing I can think of is that this is a pure tech demo to show off how clever they are that will never make it into production at all. Which is fair enough I guess, it certainly achieves it's aim if that is the case.
https://www.bbc.co.uk/news/av/business-38897417/the-ocado-wa...
There's always something mesmorizing about them!
To be sure they are "cleaned up" but the way the counterbalance is slung underneath, and "Wheeler"-like ("Return to Oz" reference) mobility, single arm with some kind of suction gripper .... it screams utility.
I bring this up as warehousing is all about reducing costs from all angles.