I'm a newcomer to arduino and robotics and am struggling to find out how to control at least 12 actuators. I'm looking for a solution that maximises strength and minimises cost. At present the physical parameters are wide open but to give an idea of my ambition it'd be great if each actuator could deliver the mechanical equivalent of lifting and lowering a few kilograms over a distance of 20cm or so in a couple of seconds. Precision might be nice but I can probably live without it as it's a machine learning application with lots of scope for fuzzy ways of using visual and proprioception feedback to keep itself loosely calibrated.
I'm interested in recommendations of what sort of motor/stepper/servo/etc will deliver a good strength to cost ratio and how to control that many actuators. If delivering that much strength is infeasibly ambitious then ideas about how close I could get would also be very welcome.
p.s. If you're wondering what it's all for see opendrawer.org
zoomkat:
search youtube for combos like "arduino plotter", "arduino servo draw", "arduino servo arm" similar to to see what others have built.
Thank you. I've been doing lots of searching and not finding much with so many motors, most have rigid joints and total only three or four servos. The most distinctive property of the robot I have in mind is that all its joints are gimbals held in tension by elastic muscle groups. That doubles the number of motors per degree of freedom. Add to that that I'm interested in coordinating arm, posture and head movements and I need 12 in total for the design I'm working on now. It'd be a little like a much simpler and cruder torso working similarly to the amazing but very expensive eccerobot.org.
Would it be possible to use 12 little one-motor drivers like the EasyDriver for example?
edburton:
The most distinctive property of the robot I have in mind is that all its joints are gimbals held in tension by elastic muscle groups. That doubles the number of motors per degree of freedom.
Have you thought about using pneumatic or hydraulic actuated McKibben muscles?
They are fairly easy to build if you want to homebrew your own:
Of course, control will be more complicated; you need a pressure source, valves, etc - plus you need to come up with your own method of feedback (to close the servo loop).
You might also look into the links here:
...this site details many historical instances of pneumatic/fluidic designs (including the McKibben muscle, which has a fascinating history); you may find it useful in your designing.
zoomkat:
Is the below what you are trying to create?
I know Patrick and Paul well. Paul's not at all what I'm making. I'm thinking of a much simpler version of eccerobot.org with only one arm and no hand. It's the pairs of elastic muscles (like a biological arm) that are important to me.
cr0sh:
Have you thought about using pneumatic or hydraulic actuated McKibben muscles?
Air muscles are close to perfect for my application but my hunch is that compressor+valves will be too expensive. I'm aiming for a design that an hacker/artist can download, cut and just add pc/arduino/motors/camera for under £1000
Cool, I might have just answered my own question. Does EasyDriver+High Torque Continuous Rotation Servo look like a working combination? Do you think I'll be able to control the speed of all 12 servos (with my priority being accurate timing rather than very accurate position)?
(I've plenty of mechanical ideas for using springs, strings, counterbalances and visual feedback for the elastic behaviour with the Arduino focusing on controlling the 12 pulleys)
