Society of Robots - Robot Forum
Mechanics and Construction => Mechanics and Construction => Topic started by: azy on January 01, 2011, 05:02:10 PM
-
Terminator size / style arm
I decided to go with blade bones rather than box section. It was a close call but blade arms are easier to make and allow a lot of boltonability
Ive opted to use solenoids, maybe with some servos thrown in for fine control. Servos & compressed air suck noise wise , same as geared down motors.
Solenoids are quiet fast ( realistic in action) cheap to diy & can be easily mad to fit any volume you wish.
Holding current is high and they give of a lot of heat. But a big plus is they are not damaged by a fall or forced movement. This factor + the quiet sound + natural movement tipped the scales
Ive stared a wiki
http://asimov1.wikispaces.com/ (http://asimov1.wikispaces.com/)
-
You can't expect anyone to take you seriously when you're showing your 'terminator parts' on a bed.
-
lol , i only started building it yesterday, gimme a chance ::)
It will prolly take me 6 months all told. Ill post updates as i go along. This will give people a chance to chip in and also see how and why im using the techniques that i am.
The tendon is directly inline with the bones in the arm. As the solid tendon is contracted the angle changes to the most direct path. The solenoids are clustered both sides of the tendon, forward and rear. This avoids collision with the tendon
The solenoids have about 1 liters of volume available to them with out looking bulky
Additional solenoids (a second smaller bunch acting on the middle of the tendon ) can be added later if i need more power
I need a total lifting power of 1kg to raise the sprung counter balanced forearm arm
For every 1 kg of hand pay load i will need about 5kg of lifting power from the solenoids (force times distance )
Ideally you would want 20 to 50 kg of pull from the solenoids, if i get 10kg in burst mode ill be happy.
Your biceps generate an incredible amount of pull, a rough glance we are talking 300 kg !!! that's just an insane amount of power
Walking aside , the elbow joint is the most difficult. Speed and power are required , the weight of the arm also has to be overcome and the leverage of between 4 or 6 to 1 is working against you all the while.
Shoulder joints can be powered by huge motors hidden in the torso. Where as the bicep area of the upper arm is the only volume available to operate the elbow joint
The solenoids may have to be water called as they will need to be pushed well beyond sustainable limits for short bursts.
Thankfully we don't spend all day waiving our arms about and we tend to carry our arm low most of the time.
In burst mode the solenoids can easily generate 5 times the sustainable lift, handy for throwing a ball across the room or lifting a heavy weight to that part of the arc of movement where the leverage forces working against us are minimal (getting over the hump, like the lift and jerk technique used in weight lifting)
if all goes well i should be able to lob a bag of sugar with some force across the other side of the room or raise a 1kg hammer above its head
Its not designed for lifting heavy weights , im just trying to get into the ball park of human abilities of speed and power when working in everyday light duty environments (and do so with out the horrendous din of servos or pneumatics, these blast out sound right in the vocal range and are incredibly irritating as well as interfering with speech recognition )
-
Just ran a quick test on the very crude solenoid. Im getting about 10grams lift per volt . (50g at 5v)
so with 20 of these crude solenoids running at say 24v = 20solenoids x 24v x 10 grams = 4800grams = 4.8 kg
now i will certainly be able to improve these crude solenoids by a factor of 4 if not 10 , and there is room for about 72 of them ( 2 x 6 by 6 array)
so if i run at 36v x 72 solenoids improved by x 4 x 10g per volt = 103680 grams = 103 kg of tendon pull divided by leverage of 5 = 20kg dead lift at the hand
so realistically ill will be able to lift at least 2kg at the hand (which is what i want) and maybe up to 10 or 15 kg
the current draw will be high , but im using a very large car battery that can deliver 100s of amps in short bursts
the general rule for solenoids and electric motors i find is 1kg of lift/force for 1kg of motor weight
so i will need a 5kg bicep to lift 1kg at the hand
servos lift far more than their weight but they run much slower especially under heavy load and are easily damaged if forced
-
Power lifter design i may well incorporate my power lifter design.
Its a realy clever little mechanism i designed where a large motor in the trunk actuates the elbow joint (combining a belt drive and a winch mechanism) It does complicate the shoulder action. as both the shoulder and elbow have to be actuated together to cancel out unwanted forces or movements in order for the remotely actuated elbow joint to move in a typical fashion
Power isn't an issue when you have the whole volume of the trunk to mess around with.
With a bit of luck i can get my pure solenoid set up past the winning post. If not there are other ways like my power lifter design than can boost strength considerably
-
Where are you getting the funding for all these solenoids you will be using? Solenoids aren't cheap and 72 of them will break the bank on any normal non-corporate funding.
-
i should add that ive been making and testing solenoids for years. Including a 10hp solenoid motor
DIY solenoids are cheap and easy , and loads of fun top play around with (hit you tube for details)
Its just a coil of wire wrapped round a hollow tube with a steel/iron rod add a battery and the rod gets sucked into the tube.A spool of laminated copper wire will cost about 10 quid, use a drill to help with the winding.
the solenoids im building will cost about a pound each and take about 5 mins to make
ill pick up a bit of funding as the project progresses, time not money is the biggest investment when building an android. Im only building an android so the brain software has got a proper body to learn and explore the world with. Mind and body are part of the whole. I don't actually have any interest in robotics as such. Building this android is a have to rather than a want to. But its an enjoyable challenge all the same.
Im very good at problem solving, give me a problem and the chances are ill come right back with 10 solutions
-
You need hydraulics to build such an arm:
(http://www.luxurylaunches.com/entry_images/0109/05/endoskeleton-arm.jpg)
or better yet this one:
(http://i184.photobucket.com/albums/x177/NoHumorMan/Terminator/T1-Arm2.jpg)
follow the talk here (http://www.therpf.com/f9/metal-screen-accurate-terminator-arm-92767/)...
-
That terminator arm is sexy.
Cos my arm is going on an android hydraulics are out of the question Too noisy & i don't have the kit or skills to use hydraulics.
I may use some low power (quiet) hydraulics or pneumatics but it will be very different from stock setups and just used for minor loads
Part of my inspiration for the solenoid biceps came from the piston actuators on the terminator arm
On my you tube channel there is a video (link) to a full sized hydraulic/pneumatic terminator its realy cool but noisy as heck
-
I have seen some legs prototype in Italy 4 years ago (I should say 5, since we are in 2011) that were pneumatic with PWM actuation and yes, they were noisy. The problem was PWM-ing the valves to get the right pressure for the pistons to hold their position under load. I guess the same applies to the hydraulics, if you decide to use PWM. However, I have seen 11 years ago in Romania proportional control of hydraulic systems, where a valve that is opened more or less would let little or more oil to get in the piston to hold it's position precisely. And that was NOT noisy. The only noisy part in the hydraulics system is the motor that powers the pump. Hydraulics are messy, not noisy. And you get the most power in a tiny actuator.
Alternatively, you can use linear actuators (http://www.trossenrobotics.com/linear-actuators.aspx?a=gad&gclid=CNuQ_czOm6YCFUS8Kgod-RKgoA).
Or you can always convert your servos to linear actuators, like here (http://www.societyofrobots.com/robotforum/index.php?topic=10043.0).
The problem with the solenoids is that they act like a digital output, open or closed, nothing in between. I fail to see how that will be useful.
-
Hi,
The problem with the solenoids is that they act like a digital output, open or closed, nothing in between. I fail to see how that will be useful.
Just add PWM.
Fuel injectors are PWM'd solenoids.
-
that hanging monkey was a great leverage test
thanx for the link to the actuators (i have not ruled anything out as yet)
If anything im also trying to achieve something that hasn't realy been done before. I tend to look at what has gone before, and if it has issues to try and overcome them.
Quiet + fast enough + strong enough to be useful + damage proof (and hopefully some level of control)
It is the damage proofness of directly acting solenoids and their quiet actuation that is strongly attracting me ( at the moment im willing to accept huge current draw and excess heat) my android wont be working 8 hour shifts down the saltmines. Perhaps acting out a demo lasting a few hours.
ref Just add PWM . Yes i agree there should be a number of methods available to pulse the solenoids or operate them in batches.
i may add some springs to some of the solenoids to A give me some elasto jerk stiction breaking power and B convert a stalled high current drawing heat radiating solenoid to one that gives a gentle tug throughout the raising of the forearm
this is another way to turn a square wave into a sine wave, by physically buffering the load the coil has to deal with. Reducing contact arcing & hi freq esc buzz.
Thinking out loud
I think some kind of flash steam electro heated pistons encased in vacuum tubes to avoid heat loss may have a future, high power to weight ratio and quiet....if i had the money & time that's where i would place my bets. Im thinking combining the power of an internal combustion engine with the smoothness of pneumatics while blanketing out the noise via controlled rates of piston expansion, thus avoiding pops....if you get me. I will carry out a test on this idea later on,
but for now im looking at traditional servos n solenoids geared motors etc to get the job done
electro pneumatics and electro hydraulics that are quiet , cheap and easy to use are thin on the ground
I wonder if you can wire in a capacitor with a solenoid to provide a millisecond inrush current eliminator ? my electronics gets a bit thin this deep. Kinetic and electrical shockwaves are wasteful i know that much
-
Bit of maths
The elbow joint can contract and extend (1 complete cycle) at 150 rpm... providing sharp edges such as tables are avoided :P
That's 10 times in 4 seconds or 2.5 full contractions and full extensions in a single second
hand from thigh to shoulder and back to thigh = 1 cycle = 1 rpm in this analogy
From my elbow to start of wrist is 20 to 30cm depending how you measure it so we will take 25cm as a midline. So we can work out how far its traveled and how fast and have some kind of guess at the acceleration /declaration rate etc
The end of the forearm where it joins the wrist travels about 50cm to 55 cm (thigh to shoulder) so it travels at about 2.5 meters a second. 2.5m/s x 60sec x 60min / 1000m = 9kmh = 5.6mph This sounds about right as limb contraction x limb contraction would be 30mph ish the speed of a 100m sprinter (who's unloaded and longer legs in combination can move about 50 mph)
So the speed of human limbs combined movement is somewhere between 20 and 60 mph when you include deceleration to go back the other way.
In the arm test i did, deceleration is almost instant and the recoil,bounce and tendon spring effect creates massive acceleration forces from near zero. Dynamic biological movement cheating like this means we are only half as strong or quick as you might think think. Static dead lift testing single joints will show the difference between dynamic and passive abilities. ASIMO could run faster still if he was a little smarter and a little braver and made full use of the cheats already available in his particular body.
The fastest cricket ball recorded was 100 mph. Forward motion + legs + whole arm + wrist + torso + other muscle that can be used. so if you knock off 30 mph for running speed and ignore the light load of the ball you have 70mph speed for the combined upper body movement.
This will provide some benchmarks of power speed and leverage of the human elbow joint
right now most robots look like the bionic man ...without the bionics ::) 1 m's limb movement looks pretty lame in real life
so the minimum target for limb realistic movement needs to be set 2m's
-
bit of Engineering
Now any sensible engineer (including mother nature) designs & builds with a margin of error and safety in mind. So if a minimum target is 2m's limb speed we should at the very least aim for 3 and build for 4.
As we are on a tight budget limited engineering resources , and working in an area of great unknowns we should realy be building for 6 or 8 m's capability or dare i say 10m's capability. (if you are 3 times over engineered and it still breaks !.....you are no longer an engineer)
Note: Speed is far more important to my project than strength , with ample speed there will also come some strength
As a minimum lifting capacity of 1kg is also required this gives us another set of figures to plug in
Absolute Minimum real world requirements 1kg weight lifted at 2m's by the elbow muscles alone
ideal real world performance would be 2.5 kg at 3m's
so we should build for 2kg at 4m's if we want to be sure to hit our absolute minimum target
We are willing to drop real world lifting from 2kg to 1.1 kg if we have to , just so we can be sure of exceeding the 2m's speed target
Less challenging projects will require less planning , and dare i say it procrastination , lol. Its just simply the case that "overcoming the build space & leverage issues " that accompany humanoid robotics does require extensive head scratching from time to time.
-
You should also look into air muscles. There is the Shadow Robot Company (http://www.shadowrobot.com/airmuscles/overview.shtml) who is researching this for years and they have been successful in creating a hand and lately legs. See this site (http://www.vrealities.com/shadow.html) for more data on the hand. It has the elasticity of the human muscles that you need when building an android.
Other experiment that uses the elasticity of the limbs is the MIT's COG robot (http://www.ai.mit.edu/projects/humanoid-robotics-group/cog/capabilities.html). They use electric motors with a twisting coupler that transfers motion to the limbs. The robot can flip-flop a springy ring or play drums.
So whatever you decide to use, you need to use a springy actuating mechanism to get results similar with a human.
-
already exchanged emails with shadow hand. Not interested really into their air stuff, but the hand itself looks quite good. Im in the uk and so are they so we may get something going together
AIR = noisy = non starter for me , same goes for hydraulics
AIR & HYDRAULICS are simply nonstarters for mobile robots , but they work well in industrial or animatronics applications
noisy expensive complex messy heavy need huge power etc etc etc , its not an issue with static installations
Im more after a slinky cyberwoman look a bit like this. Something with fluid movement that has a touch of magic
when it comes to the ASIMOV1 android im using a whatever it takes approach something that raises the bar. I need to build a Showcase for the brain , one that doesn't suffer from the usual robot issues such as noise,jerky movement, uncanny valley , dumb valley over sized hands, grating vocals etc
Phillips "Robot skin" (LONG VERSION) by Aveillan (http://www.youtube.com/watch?v=b0SVIA5SaLg#)
-
This is a pretty good form for a masculine bot (size 20 boots aside)
as robo mania is sweeping the planet a number of offerings are starting to hit the mark style wise.( at last) Enthusiastic home built diy androids are also taking their first steps. Net books are lightweight computing power houses & can handle more general multitasking roles , while plug in boards get down n dirty with the mechanatronics. Walking Androids are never going to be cheap but the cost is measured in thousands not tens or hundreds of thousands , and some highly creative efforts have been built for 3 figures sums. Junky yard ASIMOS are starting to become a real possibility if you are creative and determined enough.
I think over the next 15 year's we are going to see Android's realy coming together.
-
The new 2010 HRP 4 has a slim frame and can walk at relaxed human speed pace ( the C version is the female released earlier )
Typical run times for current androids are claimed to be around 10 to 20 minutes , which sounds about right from the longer demos, and perhaps and 3 to 5 minutes run time if walking. This would be inline with some of the high performance RC models ive seen and built.
-
Here are some of the lipos ive been testing, note the thick wiring , gold plate banana plugs and the balancing lead. I use even thicker 9mm silicone wire elsewhere just to be sure.
Lipos give you the power to weight ratio you need for androids, controlling that power gets expensive real fast as the ESC (speed controllers) are not cheap. I plan on combining simple on/off switching for the heavy loads where you need 100% power bursts with smaller ESC's for fine control of the legs & arms, this will help greatly reduce the cost and complexity.
Having a 7ft android , needs more power but overall its more efficient. Having plenty of working room with over spill makes life a lot easier with tricky high power projects. Ill use a large car battery for testing , but ill have to cram it with lipos for walking any kind of distance.
-
Just done a lifting test with diy linear actuator (aka threaded bar + 4 nuts + encased in box section metal + cordless hand drill)
I managed to get about 10kg's of lifting power, translated into about 2 kgs actual lift at the wrist
TIP , there is sideways torque , 2 actuators turning in opposite directions will cancel the forces out ( this is not an issue if the actuator is bolted to a bench, the bench isn't going to rotate unless you have the mother of all actuators
Downside of using actuators is that they are very very slow , mine was retracting at around 1 cm a second, translating into about 5cm a second at the wrist. It would take roughly 10 seconds to raise the forearm from thigh to shoulder.
-
Just FYI. Your reasons for not using hydraulics/ pneumatics(air) is that they are noisy, bulky and heavy. This doesnt have to be the case. You could just design the system well.
Most of the noise/weight/bulk behind these systems come from traditional combustion engines used to power/pressure the system. Maybe if you were to use electronic motors to do this, you would certainly be removing most of the noise, a vast amount of the weight and bulk too. You probably wont get as much force(or not at the same availability) but again its all about how well the system is built.
Pressures levels can run the same as gearing, so a small motor can generate more torque by gearing whereas in pressure, a small air pump can gain more torque by pumping air from a small cylinder into a larger one.
Also with these systems, you can control how much pressure is used to allow a full range of movement. Solenoids on the other hand Normally go from 1 extremity to the other, especially with homemade ones. You are likely to find that the only movement range that you can really accept with your solenoids is the arm completely bent or completely open but not any range inbetween.
-
Ref air or liquid systems ,Yes they could be powered with electrics, but the systems would still be complex power hungry and bulky compared to 100% electric
If you have ever used a simple electric tire inflator , they make quite a racket. A quieter system could be manufactured but the cost of this alone would run into the 1000's and require months of R&D there are no cheap & quiet battery systems on the market capable of generating 10hp . The same money would be better spent on lithium cells or other parts of the project.
Key point is that my bots have to be mobile. 90% of all mobile robots do not use hydraulics or pneumatics , & prolly for the very reasons i have pointed out.
Its not that it cant ever be done ( air or liquid systems) , its just that there are low cost mass market alternatives & im following the path of least resistance on most aspects of the build.
-------------------------------------------------------
If i go for the solenoid route i will have a dozen or so just for the arm , they have a nonlinear pull, dependent on how far in/out of the coil they are. Pull can be varied by adjusting the voltage, i would need a position sensor to get them to hold the arm at a certain point. Ill prolly incorporate a dumb self regulating mech to reduce the complexity of control .
Im still testing and developing different systems , ill prolly use a combination of solenoids , geared motors and servos. Its a very tricky task to get the power weight bulk costs and accuracies in the right ball park. Im focusing on the mechantronics side of things for now, though im keeping in mind complexity and cost of control. This will be the toughest challenge yet for me. I have set the bar very high & im going to have to dig very deep to pull an effective true android design out of the hat.
I may well have to give in to a little noise as seen in HRP 4 , but i want to exhaust the silent options first. Hence the solenoid investigation
-
Your idea reminds me this one I saw from http://www.robotshop.ca/mechate-robot-hand-4.html (http://www.robotshop.ca/mechate-robot-hand-4.html).
(http://www.robotshop.ca/Images/xbig/en/mechate-robot-hand-3-B.jpg)
-
Now I call that "Beautiful Engineering" :o