Playing with plush toys via direct hand manipulation is easy, fun, and useful for telling some kinds of stories. However, at some point, the more advanced user is going to want some semblance of hands-free animation. So, that being said, let’s revisit some more general terms, that of “puppet,” and review the more advanced form typically controlled by remote strings, marionette.
20181226/https://en.wikipedia.org/wiki/Puppet
20181226/https://en.wikipedia.org/wiki/Marionette
20181226/https://en.wikipedia.org/wiki/Supermarionation
So, here’s the deal. If you want to use a plush toy unmodified for hands-free animation, you must use some method which is more similar to that of the marionette, where the strings attach on the outside. There are some puppet forms where the strings attach on the inside and are controlled from there, but of course that would require modification of an existing plush toy.
So, now let’s talk electronicifcation. This is, of course, on the subject of Raspberry Pi, so of course we must use a Raspberry Pi or Arduino microcontroller of some sort. Since these microcontrollers are so cheap and tiny, why not embed the microcontroller as close to the plush toy being controlled as possible? This means the battery pack must aso be carried close to the plush toy. Also, since we are talking mechanical movements, of course we will also be using motors and rotational resolver sensors. Finally, an important element of having such close embedding to the plush toy means that we can’t use strings for control, but must rather use a rigid armature as an exoskeleton.
So, now let’s get into some more details of this design. We have a nice overall desgin of a rigid armature with joints that is easy to copy movements off of to animate a virtual character. What are the joints that we must support for plush toy animation? For most plush toys, the joints are rather limited, so our description of the requirements will go rather quick. We will describe the animation properties cumulatively, with each more advanced level including all properties of the previous level.
All movable object animation:
-
Location and orientation of the object, as a discrete movable object.
-
Slightly advanced: for objects that are to be thrown into the air, a launcher mechanism that stays on the ground to launch the object. In practice, the launcher mechanism cannot travel with the object due to weight restrictions and the fact that it may undesirably alter the object’s inertial characteristics.
Very basic plush toy animation:
-
Ability to roll around on one’s own. This requires a counter-weights and levers mechanism to allow a plush toy to, for example, roll off its back to sit up.
-
Rotation of the head. All head rotation must be limited to not exceed bend and twist limits.
Basic plush toy animation:
- Movement of arms together as one unit. Typically this is just rotating the arms up and down if the arms are pinned together.
Typical plush toy animation:
-
Rotation of legs or feet. A single segment without a knee joint is the norm.
-
Rotation of arms. A single segment without an elbow joint is the norm.
Somewhat advanced plush toy animation:
-
Adjustment of an elbow joint at the arms. Most notably, this allows the plush toy to fully adjust their arms to “pick up” objects in a more general manner.
-
Separate rotation of top of legs from feet. Typically, this may just be a single-axis joint similar to an elbow joint, but in more advanced models, it could also be a 3-axis joint.
-
Claws at the end of the arms to pick up objects for real, similar to how it is done with fingers.
-
Claws at the end of the feet to climb ladders, sticks, and so on.
-
Pivoting of back, with bottom staying at stationary angle. The more basic form of this animation is to simply rotate the entire body, or rotate the body and rotate the neck to achieve a similar effect.
-
Twisting of the back, the primary purpose being to rotate the arms/shoulders with the bottom staying stationary. The more basic form of this animation is to simply rotate the entire body.
So, given the above specifications, what is the maximum component complexity required for an armature?
- Battery pack
- Microcontroller
- Accelerometer
- 3-axis magnetic compass
- Precise location determination beacon
- Wireless communications antenna
- Counter weights
- 6 x single-axis motors/resolvers for counter weight control
- 7 x 3-axis motors/resolvers
- 1 x 2-axis motors/resolvers
- 7 x single-axis motors/resolvers
Total single axis motors/resolvers: 36
For friction free motion recording, choose a resolver in preference to a motor. For live animation copying, choose a resolver.
Once you’ve got this physical framework defined, it is pretty easy to record motions as played out of physical models, copy motions on physical models, and, most importantly, copy motions on phyical models to virtual models. Once the base motions in the virtual models are properly animated, you can add special effects in any way you please.
One important special effect is lip sync. Compared to more advanced puppetry, notably missing from the previous specifications is mouth movement control. By all means, that is recognized to be one of the major limitations of playing with plush toys, and this limitation is imagined around in typical play. Also, this is where transfer to virtual models has the biggest advantage. With correctly synced character voice recording, lip syncing can be automatically generated and mapped to the virtual model’s mouth.
Another important effect of the digitized puppetry is telepresence. I’ve thought long and hard about this for quite some time, and it seems that the primary problem with video-game like control of virtual characters for telepresence is that it is not immediately intuitive for all users. Although joysticks, keyboards, and mice certainly do certainly help make controlling a robotic arm, issuing macro command sequences, and pointing and shooting easier, these actions are not of the typical nature of real-time puppet control. Oh, and unlike traditional 3D animation, where animation rigs are entered in key frames not in real-time, moving rigs into place little by little at one 3D view at a time, the animation must happen in real-time and be manipulated from a single view in order to be intuitive to a simple user, as ease of use is the emphasis of plush toy play over more sophisticated puppet control.
Also, there are more considerations to making this user-friendly. Some plush toys can be dressed to wear a shirt of some sort. Being able to fasten the armature into a piece of clothing is a great way to hide it during play, and also to make putting it on a bit more user-friendly, hopefully. For leg control, this implies some sort of pants may be needed. For head control, wearing a head band or a helmet of some sort will be needed.
Again, it’s important to have some sort of physical conceptual base before you can do higher-level operations really well. Finally, you can argue that it would be most convenient of all of the armature measurement could be done analytically by remote cameras mounted throughout a room. Indeed, that would be ideal, but it requires quite a bit of formal knowledge in advance in order to do easily and to do well. If you already have a 3D scan of the plush toy in question and a formally designed armature by which to map motions too, mapping motions from video cameras can be straightforward. In this case, having the plush toy wear a special shirt with symbolic markings can help the camera system transfer the motion vectors to the virtual armature.
In all of these cases, being able to quickly get an accurate 3D model of the plush toy in question is dramatically useful to expediting the process.
36 motors! Wow, that’s pretty complicated for a little plush toy. Some pretty sophisticated vehicles might only have 50 motors, which is not many more than that toy. Yeah, but at the same time, let’s put it this way. It’s about the same number of motors that your human body uses when playing with a plush toy using only your arms and hands. Also, an anthropomorphic model robot of the human body could easily end up having over 100 motors. So, for sure it’s a simplification, even though it still comes out pretty complicated.
Now, for the mechanical and electrical engineering part of the equation, let’s just hope we can say that the tools available to us have improved greatly and that it will no longer take a Ph.D. in mechanical systems motor control to build these kinds of systems.