Larger sized Humanoids

[wthierry]

If going for a larger size humanoid, such as the Asimo size robot, are strong stepper motors used, or are large servos used in that as well.

Thanks

wes

[wthierry]

The reason I am asking, is that by building one at a much larger scale, you have room to put more advanced sensors and controls, a laptop for instance can interface everything.. Im just not sure about what they use for motors in a bot that big.

[SK]

Larger Humanoids like ASIMO or Pal Technology´s REEM-B (see http://www.ouroboros.org/papers/reemb2008.pdf) use motors with Harmonic Drive reducers, the exact cost of which I don´t know, but I often hear they´re very expensive.

Robotis´ relatively new EX-106 servos might enable slightly larger humanoids at "low" (relative to harmonic drive) cost.
But from what I´ve seen in the RoboCup Humanoid TeenSize league, even the larger humanoids built by research groups around the world for this competition tend to be of rather lightweight construction. There certainly isn´t much room (and especially free payload weight) for putting a laptop and other stuff into them.

The recent most extreme example certainly is Hajime Robot 33 (at over 2m height and only 20kg weight):
http://www.youtube.com/watch?v=FSAaC8DrlQI

[wthierry]

It seems to me that maybe in some cases stepper motors might be in order to give the bot more accuracy in walking than servos, am I out of line here, or is it just the expense of stepper control drives that scare people away from these? The motors are relatively cheap at surplus houses, and pretty strong compared to servos. Feel free to beat me down, since Im new to all this, but my thought for this would be to attempt to use steppers for the legs. Of course, Im thinking big, my first home brew is going to be of the average size and use servos, but I am going to attempt to use styrene plastic as the frame material, since its much lighter than metal brackets, and fairly easy to work with and secure together.

wes

[Robo1]

I've had experiance building full sized humanoids. There VERY expensive. Stepper will not cut it. We use maxon motors that are $800+ then you add the harmonic drive $2,000+. There also very hard to control. The bigger the biped the harder the control problem. You also need high resolution position feedback something like 16bit + e.g. 64,000/turn. Plus a high band width to control the motors 5khz+. It took a team of 20+ phd + tech's over a year to get it up and running before we could even start trying walking and balancing algorithms. For a bot this big you can use lots of sensors but there all have to be high resolution $1,000+ IMU's, the 6 axis force/torque sensors in the feet cost $10,000+ each!

if you have a spare $150,000 then you can go for it. If you can make it work for $10,000 then you will be very rich.

Maybe start with a small boy about 90cm.

Hope this helps

[billyzelsnack]

20+ PhD's trying to accomplish something. Sounds like that's the problem right there!

[billyzelsnack]

Here's a very interesting post on steppers and servos from a guy that makes both types of motor drivers for the CNC community. I'd tend to keep in mind what this guy says because his drivers are used in thousands of very high-precision and high-torque requirement machines. I guess the difference with robotics is that we actually care about weight as well which kicks steppers out of the running until you get into the big bipeds.

http://www.cnczone.com/forums/showthread.php?t=52090

Using his formula for motor wattage that puts an EX-106 at about 60 watts. An EX-106 is probably still too small for a full size humanoid.

http://translate.google.com/translate?u ... n&ie=UTF-8

Uses a VS-SV3310 VSTONE servo which is a 327kg-cm servo. Even with its lightweight and high torque servos it still looks a bit under powered.

So it looks like a full humanoid would fall at least beyond the 200 watts stepper motor boundary he mentions. So it seems that servos might be the better choice according to his metric.

[wthierry]

So if I rule out the Full sized Biped, what about a medium size biped such as the size QRIO was supposed to be, does anyone know what kind of servos and such these bots were before they were cancelled?

[Tyberius]

So if I rule out the Full sized Biped, what about a medium size biped such as the size QRIO was supposed to be, does anyone know what kind of servos and such these bots were before they were cancelled?

Nothing commercially available I'd imagine. I've built a functional 24" tall (about the same size as QRIO, if I recall correctly) using RX-64s from Robotis, and have seen nearly 3' tall bots using single EX-106 bots function without issue. That's still $3400-$5600 USD in servos, just for the legs, however.

[SK]

According to http://www.aibo-kennel.com/QRIO/QRIO_Brochure.pdf , the QRIO is 58cm tall, for a rather high weight of 7kg. Incidentally, 60cm also is the maximum height for robots in the RoboCup KidSize league, with most robots of 50-60cm size weighting 3-4 kg. The ones you can see in my sig videos are using 3xRX-64 and 18xRX-28 at a size of 57.5cm and weight of ~3.5kg.
It´s thus a pretty good guess that Sony´s "Intelligent Servo Actuators" for QRIO are quite a bit more powerful than RX-28s and RX-64s, but probably also much more expensive (and unavailable).

[jocke]

According to 'Mechanical System of a Small Biped Entertainment Robot', 2003, available from ieee. Qrio uses three different actuators:

ISA-4S: 0.5Nm, 80g
ISA-4M: 1.1Nm, 130g
ISA-4MH: 2.1Nm, 150g

ISA-4MH is used for the legs, ISA-4M for the trunk and leg axes and ISA-4S mainly in the arms.

Their weight explains why Qrio's total weight is so high but they are pretty week, the knee servo is only around 20kg/cm.

[jocke]

Start-up torque is about 3-4 times of rated torques.

[billyzelsnack]

I guess by starting-torque they are talking about stall torque. So the 2.1Nm (21.4kg/cm) is moving torque.

If the ISA's stall torque is 3 times 21.4kg/cm.. that is 64.2kg.cm for 150g.
If the ISA's stall torque is 4 times 21.4kg/cm.. that is 85.6kg.cm for 150g.
A RX-64 is 64kg/cm for 116g.

Anyone know the moving torque of the RX-64 so that a proper comparison can be made? Given the numbers above an ISA and a RX-64 seem quite similar in performance.