Boston Dynamics Stretch

Just saw Boston Dynamics’ new Stretch robot. Looks like it’s using some sort of swerve drive (visible on SP3 at 0:55). We never see the bot rotating and translating at the same time, so it could be a simplified swerve-esque design (which would make sense) but yeah. Just thought I’d throw the video out there since my quick search didn’t find any posts about it yet.

Also interesting that they show it using a suction gripper on cardboard. Seems like you would be limited to very light boxes with such a technique.


I am willing to put my money on a turret on a swerve. Crab doesn’t make a whole lot of sense to me in a industrial setting. There are just space/positioning benefits to keeping the drive base aligned to the factory floor grid.

They have something close to 40-50ish in^2 of suction on that thing`, Industrial suction grippers are no joke. Even if we assume they are only are pulling 500hPa that grip could (in theory support several hundred lbs.

` Dead recon guess


I’ve never worked in a warehouse with heavy-duty cardboard boxes, but I could imagine trying to hold a few hundred pounds by the side, or even the top, of the box, the cardboard or tape holding the box together would break or rip. This wouldn’t necessarily happen when supporting the box from the bottom.

I’m sure Boston Dynamics has the right solution to their problem though.

Me neither, best i can relate is road cases for concerts. Cardboard boxes rarely weigh several hundred lbs and would never be that heavy stacked like that in the demo. In the instance of heavy contents it would be a plywood crate, or at the very least on a pallet, moved with a forklift or pallet jack.

However, that gripper has a lot of surface area, and it is spread out quite a bit. I suspect dropping the box is not much of a concern. Cardboard is not particiuarly permable to air, so it is fine to pull a decent high-flow suction against. And they did pick up an ~70lb SPOT robot (albeit a different surface to pull suction on)

Pick and place robotics with suction has been around for decades at this point, there is probably a good chance that gripper is COTS anyway.

1 Like

Over 100 lb of aluminum says otherwise. You get enough suction pads, it’ll go right up. (6-8 IIRC, large and constantly pumped, more in some cases.) Just don’t put a gripper on a hole.

Oh yeah for sure. Definitely not a crab drive. Just not sure if it’s a “true swerve” (which I’m thinking of as each wheel independently steered and powered) or if (for instance) the left and right wheelsets are maybe driven by the same motor allowing swerve translation and swerve rotation, but not translation/rotation at the same time. In an industrial setting it really would be functionally equivalent, esp. since there’s a turret as well

NIOSH equations for workplace safety assume 51lb/human, and packaging designs tend to respect that and target 50lb box weight limits. More than that and it generally gets palletized.

Clever arrangement of cardboard, foam, tape will easily take you into the low 100’s of lbs of weight bearing, as long as the packaging design is conducted knowing that it will be lifted by any surface. Most of the time that’s not necessary, but Packaging Engineering is it’s own subfield (& often, department) for good reason.


Hmmm, there is a hard limit of how much vacuum you can draw: it’s 14.7 psi at sea level.

Now I’m imagining working inside a pressurized warehouse where all the suction grippers are super effective…


If you look at about 0:49 and count the pads, it looks like 18 separate suction pads on the head. That could translate to a very good grip and a lot of forgiveness for uneven surfaces or small holes. My bet is that this thing can handle well above 100 lbs.

Also, I’m also in the “true swerve” camp on this one. Watching the wheels as it rotates makes me think that they are probably independently powered.

1 Like

CD is baffled that suction works with the right equipment…


That’s why i specified a 500 hPa pressure differential, it seemed reasonable to me at the time. Still: 7lbs*40in^2 is 280lbs (all ballpark guesses of course). that would have a safety factor of over 5 for that 50 lb box… so real world numbers could easily be less and the “100 lb target” could easily be hit. shrug

Overarching point is “big ol’ suction head” is perfectly acceptable for “ye ole box”

CD was baffled suction worked in 2019 lol. I’ll trust the Boston Dynamics and Pick-and-place suction head engineers on this one.

Hmmmm, interesting thought. I am sure it has been done in some weird instance/ industrial process. perhaps if something needed manipulation in an autoclave or something ? (not my wheelhouse so…)

1 Like

My bad, I thought I read 40-50 lb/in^2

1 Like

So, I’m going to have to point out that I routinely pull 22 psig vacuum. Small pump, degassing casting compound, small container.

That said, I’m fairly sure that there’s a reason that the gauge shows 22 PSI (vacuum) rather than 14.7 PSI (vacuum). I’m trying to remember what the big lifters I used to play with pulled, but I think it was something less than that, might have been closer to 13/14.

You either live 2.2+ miles below sea level, or…

Does your gauge look like this? This is commonly called a “30-30” gauge, which reads in units of psi above atmospheric pressure, and in units of inches of mercury below atmospheric pressure. I think the ultimate pressure of your vacuum pump is actually 22 inHg, which is a reasonable ultimate pressure for an inexpensive pump.


1 Like

When it picked up Spot and imitated a human


Nah, it’s a full suction gauge, no positive pressure readings of any sort. I’ll have to double-check the units.

Closest McMaster is McMaster-Carr so I’m guessing you’re right and it’s inHg instead of psi. If it’s psi when I check in the morning… well, I’ll have to file an enforcement ticket with the Laws of Physics. :wink:

Edit, the next day: inHg. The gauge also has bar but I generally ignore that.


It is very impressive, especially since they have it all modelled so that one can “try it out virtually” before executing in the physical world.

Please keep in mind that this is only a demo. No doubt they have designed Stretch to work with other types of manipulators.

1 Like

Assuming the pods can each rotate independently of each other (which the video shows they can), rotation + translation is 100% a software choice at that point. My guess is that it is much simpler / safer to have the robot move, rotate, then move again.

Or perhaps they actually are rotating while translating in order to maintain a heading, but the chassis doesn’t drift far enough for it to be noticeable. Super cool either way.

This topic was automatically closed 365 days after the last reply. New replies are no longer allowed.