Disney Imagineering’s “Project Kiwi” Bears Groot

Some days, we might be forgiven for believing Boston Dynamics has cornered the market on walking robots. They (and other players) are making incredible progress in their field, but three years ago Disney, trying to create autonomous, free-walking robotic actors for some of their more diminutive film characters, found none of the existing platforms were appropriate. So they set their Imagineering department to work on “Project Kiwi”, and we are now seeing the fruits of those efforts.

Research on bipedal robots has amassed over the years, and as the saying goes, if these Imagineers saw further it was by standing on the shoulders of larger robotic platforms. However, the Project Kiwi designers have made a laundry list of innovations in their process of miniaturization, from the “marrow conduit” cooling system which forces air through hollow bones, to gearing that allows actuators to share motors even across joints. The electronics are distributed around the skeleton on individual PCBs with ribbon flex cables to reduce wiring, and almost every component is custom fabricated to meet the complex size and weight requirements.

Even in this early prototype, Disney’s roots in life-like animatronics are evident. Groot’s movements are emotive, if a bit careful, and software can express a variety of personalities through his gaits and postures. The eyes and face are as expressive as we’ve come to expect (though a keen eye for seams puts off some definite Westworld vibes). Reportedly, this version can handle gentle shoves and contact, but we do spot a safety cable still attached to the head. So there’s probably some way to go before we’ll see this interacting with the general public in a park.

Disney’s Imagineering department has been doing some amazing work with robotics and they continue to make significant innovations in the more traditional fields of animatronics. It certainly looks like one of the coolest places to work right now, and now we’re itching to build our own bipedal friends to play with.

31 thoughts on “Disney Imagineering’s “Project Kiwi” Bears Groot

    1. The motions are a little bit too precise to not be CGI, because it takes quite a bit of force to fight inertia and throw a realistic punch. Either they have motors that nobody else in the industry has, or they’re still faking it.

        1. Rubber bands would have the opposite effect. What you need is a motor that a) light and narrow, b) has no wind-up, c) has high torque and power directly off the shaft

          There are disc-shaped motors that can do that, but none of them would fit inside “groot’s” arms here because they’re the wrong shape. Therefore there must be some trickery going on here, some mechanical compromise or jiggery that makes the movie magic happen.

      1. Look up a video called “A hybrid hydrostatic transmission and human safe haptic telepresence” from DisneyResearchHub YouTube channel.

        The movement are really smooth (yes still controlled by a human but the motor is capable to generate a smooth movement)

        1. That certainly is really smooth, but it’s because it appears to be directly hydraulically controlled and powered by the operator – there’s no motors or sensors involved (other than the vision system). Effectively a hydraulic version of a pantograph.

  1. I can’t express how cool this is, but yet I’m missing something very important.
    Although many people would claim that it ruins the experience it attempts to create… I really want to hear those motor sounds. All those up-beat-inspiring-uplifting-music is fun, but let us hear those motors.

    Nehhh… just like any other commercial, no real sounds are played. Maker videos with flip-dot displays… no sound, vacuum cleaner commercials no sound, electric toothbrushes in commercials make no sound (but in reality to virbrate the teeht out of your mouth, electric vehicles driving 50mph and claim to make no sound… why (they still have wheel don’t they)? To me true sound would make it much cooler and much more believable as a robot.

    1. “true sound” is a Hollywood special effects fiction. “true sound” is ambiguous and unhelpful. The sound you hear in movies is totally fake and unrealistic, because the “true sound” is so unconvincing, so unhelpful, that they have to edit it out

  2. Basically, if you have a 1″ rotor spinning at 10,000 RPM, weighing about half a pound, what you have is roughly 10-20 Joules of kinetic energy depending on whether the mass is distributed as a cylinder or a tube. Why does it have to turn so fast? Because otherwise you don’t have enough power density in a small motor. Motors at this size really struggle to make any meaningful torque, so you can’t use them without a reduction gear.

    In order to make it stop and reverse motion in let’s say 0.2 seconds, you have to apply something like 20-40 Joules of energy to cancel out the momentum and accelerate in the other direction, which at the face of it takes 100-200 Watts, which is probably a significant portion of the power it has on offer, but also makes the gears go ka-crunch and the whole thing spins around its axis from the sudden acceleration. That would make the robot twitch and shimmy around, which is no good.

    So, what the gimmick must be, they’re running the motor at a constant rate through a compact reduction gear and using a friction clutch to throw it between forward and reverse, slipping the clutch for torque control. Of course the clutch packs will wear out in a few hours, but that’s how long they need to work anyhow.

    1. Muscles are amazing like that. They just pull directly and generate hundreds of Newtons and Watts without any wind-up required.

      Compare and contrast with the mini cheetah robot:


      It gets high torque directly by using a disc shaped motor with a relatively large radius for leverage. Basically a high performance outrunner BLDC. There’s two of them in the joint – the second motor runs a belt down the leg for the knee joint. However, that’s the wrong shape to fit inside the slender arm of a “groot”.

      1. Compressed air can do that too, a lot of the very best realistic bots seem to be air or hydraulic. seems like what we need is really cheap proportional valves, micro-compressors with active noise suppression, and some really good control theory.

        1. Emphasis on compressed. Plain pneumatics is sloppy, so you need a sort of differential system where there’s constant pressure going both ways, and the difference in pressure is what drives the actuator. This is a little more complicated, but otherwise you get the “wind-up” problem again.

  3. Back in 1989 (1988?) my cousin and I were watching a show about the making of “Who Framed Roger Rabbit?”
    Upon seeing the mechanisms used to have the “Toons” interact with the physical world, she said something like; “It’s a shame they aren’t using their knowledge to improve life for disabled people.”

    1. ” “It’s a shame they aren’t using their knowledge to improve life for disabled people.””

      How do you know they didn’t? Knowledge is portable. Engineers move around, and talk, and publish. What is learned making movies, or rockets, or Large Hadron Colliders percolates out into the world.

      That said: hacking together an apparatus to pull off a shot in a movie is a long way from engineering a device for every day use in the real world.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.