Different robotic joint designs (fast/high torque/simple to manufacture)

What are best designs worth trying? Some starting point. Maybe few insights from someone who tried to build such joint and any interesting materials (patent/video/etc.). Interested in leg/arm joint that achieves performance close to SpotMini.

As I understand high torque fast motor is a problem, bulky, high consumption, etc.
How about combining many smaller motors? Distributed array of small motors arranged in 3d printed joint, individual motors may be activated depending on torque.

At first glance, it does not seem like an adequate solution; from the physical point of view, smaller motors would be less bulky and therefore would have less mechanical resistance. Keep in mind that there will be the same torsional force at any cross section of the arm. Moreover, in having more moving parts and more devices to be controlled, this leads to a larger quantity of items to have problems and/or need maintenance.

Any information on how SpotMini joints?
According to this

(https://iit-dlslab.github.io/papers/khan15sicfp.pdf)

similar robot uses hydraulics:

Hydraulics has the significant advantage of being able to absorb high impact loads (no gears required), having a high power to weight ratio and allowing the implementation of high performance torque control

Click to expand...

But having an on-board hydraulic power pack is a burden to a legged robot. It is not only increase a weight of a robot but also raise the joint torque requirements.

Click to expand...

But here is electrical motor solution, looks very similar too:
https://www.robotdigg.com/product/1667/MIT-Robot-Dog-high-torque-joint-motor

An automobile window motor or windshield washer motor, 12V 1A.

There is also the idea of a power source running through the entire machine. Limbs can tap into that power via valves, or solenoids, etc.

Factories once had a long driveshaft running past several work stations, where each workman could attach a large belt to take power from the driveshaft and run his machine.

In the same way you could pressurize air (water, oil) in a long sealed tube. The tube is installed close to all moving joints. By opening valves, this pressure is available to inflate a bag (or extend hydraulic cylinders, etc). The force then pushes against mechanical levers and cantilevers.

Air can move rapidly through a hose, creating quick movement of limbs.

An additional hose with vacuum (low pressure) might be useful for creating opposite movements.

You've also got the issues of control and back-lash.

Surprisingly, you may do well with high-end 'modelling' servo motors. Trivial to control, well documented performance, tough as hi-tensile nails, and a truly astonishing variety of power / weight / speed / cost to choose from. Also, your mini-me prototypes may be readily scaled...

One factor is 'creep', where tiny errors accumulate on an arm or limb position. If you cannot include a 'reset' pose, you may need an optical 'beacon' or two, literally a narrow-beam laser or LED on the arm, and several photo-diode quadrant position sensors at strategic positions in reach, allowing for error measurement and correction ...

Alternative is well-geared stepping motors, with optical position feedback to be sure, to be sure...

==

Been a while, but I 'played' with screen-wiper and window-winder motors for crude 'animatronic' models. IIRC, they are so customised for cars, they are a nightmare to interface with your own hardware. Trad, analogue Wiper motors are designed to be most efficient 'forwards', only reversing to park. They may have several parallel windings for the differing speeds. Threads are minimal, often a weird size, and you may have to get LH / RH pairs. IIRC, I had to open up and add a ground lead to the screen-wiper's connector box, cut a longer metric thread on the shoulder of its shaft, use the existing nut as a lock...

YMMV.