fun 4-legged robot

[Admin]

Its not very effective at walking, but its still really neat looking.

[youtube=425,350]JnfEyMRDUHk[/youtube]

I think that if it had rubber feat that it would work much better . . .

[oelewapperke]

Looks simple enough to construct.

Seems like it's just 8 servo's attached in a pretty straightforward way.

This thing would benefit a lot from feedback loops instead of straight overpowering. Can you get a force measurement from servo's ? Ie. can a servo report how much resistence it is encountering ?

[oelewapperke]

Is there any way to get a servo to report the resistance it's encountering ?

If I remember my motors class you basically measure how much juice the engine is sucking up, but does anyone have an idea how to do it ?

[dunk]

i don't know how accurate it would be but you would get some idea by measuring the current the servo is drawing.
it would be difficult to determine the difference between a moving servo horn and a stationary servo horn under pressure.

maybe if you combine readings from an encoder on the servo and current draw....

dunk.

[JonHylands]

This (the issue mentioned by dunk) is actually a pretty big deal. The AX-12 servo (part of the Bioloid robot kit) has this type of measurement, and you can't tell whether the current draw is due to external forces on the servo, or just the current draw of the motor as the servo is moving, or both.

Series elastic actuators actually solve this problem nicely - the force measurement is independent of the motor current, so it is always measuring the external force.

- Jon

[Admin]

This thing would benefit a lot from feedback loops instead of straight overpowering.

I guess the reason why I pointed this bot out is that I really liked the various gates he used . . . He just needs a global sensor to give the bot some direction.

Can you get a force measurement from servo's ? Ie. can a servo report how much resistence it is encountering ?

That would require a force sensor on each servo, an expensive and difficult task. The method Ive used in the past, although not entirely accurate, is to use a current sensor to measure power consumption.

[oelewapperke]

So one would have to measure the current, along with the encoder in the servo. Should be doable I guess.

A servo has a built-in encoder, but is it possible to read it out ?

[Admin]

A servo has a built-in encoder, but is it possible to read it out ?

Ill probably be proven wrong on this, but to my knowledge no hobby servo has an encoder. They all have pots inside. If you want to read the servo angle value, hook up the pot to an ADC on your microcontroller. Some servos offer serial feedback, and the Hitec robot servos somehow (I dont understand) have the ability to report angle information.

[dunk]

i've posted this link before but it's relevant here:
http://openservo.com/
http://www.societyofrobots.com/robotforum/index.php?topic=928.0
this project reads the output from a servos pot and measures the current to the motor.
i'm not sure how much work has been done to the firmware to integrate the 2 results into an estimated force though...

dunk.

[oelewapperke]

I like this because I've programmed kuka bots.

Force-feedback is what makes these robots execute "human" maneuvres. If you directly steer the position of the robot they shake, they "force", they're unable to pick stuff up because they exert so much force at the least bit of unexpected reaction that whatever they're grabbing crumbles. In short, they're stupid, shaky machines.

With force feedback, you can do neat tricks.

For example, steering the expected downard force will allow a robot to balance a place (and 10 plates too, if your feedback loop runs fast enough) on a stick without it falling over. It will allow a solid steel robot hand capable of crushing a car to pick up an apple without scratching it (a big problem in trying to design fruit-plucking robots : the fruit has an irritating tendency to be completely destroyed by the plucking machine).

I would like to see how effective a robot would be able to walk by dividing the force vector I'd like to see for the body over the different legs (ie it would have to counter gravity, and get the body ahead some). I think (hope) a robot controlled like this would be able to run and even to jump. But you really need to have a force feedback loop controlling the legs to get this operational. And I'd like to do this for a somewhat reasonable price.

[JonHylands]

I'm planning on building a robot eventually that does just that - using series elastic actuators, it will be completely force-driven.

- Jon

[oelewapperke]

Series elastic actuators ? Are those the $50 korean servos you're talking about ?

[JonHylands]

No, series elastic actuators are linear drive actuators with a built-in spring and tension measuring sensor.

See: http://www.yobotics.com

- Jon

[Hal9000]

It's like a turtle!