Building a Biped Robot

[airman00]

I decided to build a small biped robot and then scale up for my butler robot and eventually into a biped for me to ride in.

I'm reading up on old posts and also looking at other bipeds online.

I was curious to see if anyone else that is active on the forum ( or online to read this  ) built a biped robot . If you did please post up pics


I'll post up my CAD design , all my calculations , etc. once they are completed ( in about 1 to 2 weeks I hope) . Plus I will document it heavily and make it a nice members tutorial .

,Eric

[airman00]

Quick question : which is better on a biped - 3 axis accelerometer  or pressure sensors on the feet ?

[airman00]

OK , no replies to my thread , not very encouraging....


I decided upon a 3 axis accelerometer and a gyro as the sensors.

My algorithm will take the data from those two sensors ( which is with how much force its tilting, what direction is it tilting in, and what is the current tilt angle ) and then depending on which direction its tilting certain servos are chosen and they turn until balance is reached . Also, if the force of tilt and tilt is too much it will take a step forward (or any other direction) to counterbalance it , just like a human would do if he/she was pushed.

What do you guys think of that ? I need some help with this thing.

Thanks,
Eric

[Ro-Bot-X]

I would also get the pressure sensors for the feet. They will help with irregular terain, say going from hardwood to carpet (transition). The robot takes a step knowing how long it takes. If the sensor doesn't get triggered in time, the robot has to adjust the move. Also, I would end the step move only when the sensor is triggered so the weight transfer on this leg can be properly done. This sensor also helps with stairs (proper size to the robot of couse...).

[izua]

and eventually into a biped for me to ride in.

put lasers on the big one

anyway, i'd go with at least one main accelerometer and gyro, close to its center. pressurse sensors are a good idea on feet, but why not get another one axis accelero/gyro for them too? except the killer programming, of course.

[airman00]

well my thinking is that it would automatically balance itself if its constantly running the balance algorithm , so even if the legs somehow screw up and don't land correctly when walking , it will regain balance

put lasers on the big one

Definitely... 

[izua]

it will regain balance

two words for you mister: uneven terrain
like outdoors. maybe mountaineering, heh. and stairs too. escalators would also be nice.
but i think uneven terrain is challenging enough.

[airman00]

I see what you mean....


So I need to have both an INU ( gyro + 3 axis accelerometer) and one pressure sensor on each foot?

Anything else?

[Ro-Bot-X]

Any other sensor for obstacle detection and positioning. Can't have encoders on legs so can't keep track of traveled distance. Measuring the step lenght is way more inacurate than the encoders. Also a compass is hard to use because of the proximity of the motors in the servos. So get at least one GP2D12 and a ultrasonic sensor for mapping. Also a camera would be nice. Plus you'll want to have sensors in the hands and legs for object detection.

There may be more, but I'm tired and I... need... to... go... to... bed.............

[airman00]

all i want meanwhile is a small balancing and walking robot . I'm not worried about other sensors like obstacle avoiding , etc.


btw I did some more research for the Kalman Filters. I can now say that I understand it ( like 80%)
Read this : http://openuav.astroplanes.com/library/docs/writeup.pdf

[R.O.B. 2.0]

 I don't know if you've seen this.

http://youtube.com/watch?v=VTlV0Y5yAww

If you choose to make it go upstairs make sure the algorithm is perfect.

[airman00]

I don't know if you've seen this.

http://youtube.com/watch?v=VTlV0Y5yAww

If you choose to make it go upstairs make sure the algorithm is perfect.

lmao

that was a nasty shot to the head

[pomprocker]

haha call the paramedics!!

[airman00]

Well I've been doing some more thinking lately......

I'll be building two versions of bipeds

One will be a true balancing biped robot and the other will be a "fake" balancing biped. For chives and that huge mech I want to ride in , I'll prob use the mechanically balancing one.

True balancing biped will have a total of 8 servos
Base its design on this : http://www.imagesco.com/catalog/biped/bipedal.avi

The "fake" balancing biped will look like this and accomplish balancing mechanically:

[Ro-Bot-X]

Check this one out!

[airman00]

I've been doing some more research and I decided to skip the small scale model and jump right into the large scale biped for me to ride on

It will work sort of like this one :http://www.societyofrobots.com/robotforum/index.php?topic=3602.msg27617#msg27617

That design isn't very efficient though , so I'll post up my design soon.

What factors affect the length of the step of the robot ( without tipping obviously) ?
For example can I compensate for center of gravity with a wider foot or something?

Btw I will probably also be installing some mechanism to help the robot shift its center of gravity ( for some balancing).

[steferfootballdude13]

This is pretty hardcore robotics... Have Fun

[airman00]

Update: I'm still debating between designs   

Just wondering why people don't just scale up on designs similar to this:
http://www.lynxmotion.com/images/Products/Full/bratblk.jpg

I mean its just larger motors with more torque ( windshield motors with encoders) . So why aren't people making giant robots???   

[Ro-Bot-X]

I mean its just larger motors with more torque ( windshield motors with encoders) . So why aren't people making giant robots???   

Maybe because they get more expensive?

[airman00]

I know it gets more expensive , but not really that expensive .   Think about it : replace each servo with a geared windshield wiper motor and encoder ( 30 dollars on ebay) .

I have a feeling there is something else besides expense that keeps people from scaling up , but I am not sure what it is exactly.

[dunk]

I have a feeling there is something else besides expense that keeps people from scaling up , but I am not sure what it is exactly.

the walker in the picture you attached has 10 servos which will all need to run together for dynamic walking.

i'm going to guess some numbers here...
so lets say a wiper motor has a 10Amps stall current.
so 10Amp motor drivers are expensive. 10 of them is getting very expensive.

a 10Amp stall current would be around 3 Amps when running continuously under medium load.
so you would need batteries capable of supplying 30Amps for regular walking and 100Amps if the bot stalls.
that's either a very large battery or a very expensive one.
maybe the bot could push it's battery in a wheelbarrow...


dunk.

[pomprocker]

I was looking at the cost of building one of these biped kits piece by piece instead.

the robonova-1

has 16 digital robot servos at ~$59 each, which i think there is a discount if you buy 10 at a time. but still thats like almost a grand right there.

then the brackets where like $7-$10 each

i didnt even look at how much the board costs, etc

so you could start by drawing up all the brackets required for the robonova-1 in sketchup to scale.

then find the biggest PWM motor controller you can afford, and the biggest battery you can afford, and then figure out what size the servos are where you could power 16 of them, and then just scale up the bracket drawings you made to that size servo.

I found an industrial servo here for your mech:
http://www.active-robots.com/products/motorsandwheels/industrial-servomotors.shtml

but its like $600, imagine that times 16.....almost 10 grand right there.

building a giant biped could be the cost of a brand new car!!

[airman00]

Thanks for the replies guys.

A biped robot such as the BRAT robot has only 6 motors



In case you are all wondering I would be using two wheelchair motors and four windshield motors. Wheelchair motors for hips and the other motors for the other joints.

I see what you are saying about the battery issue.  I dont think it will be that much of an issue , I mean 6 motors need around 70 amps(wheelchair motors draw more than 10 amps stall current)  total when stalling  , so I figure a 80ah or 100 ah battery ( around $150 on eBay) will be good enough for me. That means that at stall current it can last over an hour. Under medium load it will draw ~24 amps and last much longer

I am getting a little worried about the torque issue of the motors. After I get the BRAT robot kit and do some tests I'll know exactly how much torque is needed. I hope that two wheelchair motors and four windshield wipers fulfill the torque requirements. I am certain that if all motors were wheelchair motors then there would be no torque issue , but those motors get expensive and are heavy.

By the way my budget for this project is $5000 , so I am not really worried about the motor controllers .

[sotu]

Sounds neat airman00!

I dont mean to do any harm to the main subject of this post, but while were into this "big robots" i do have a question for a own project of mine ive been thinking about:

I was thinking of making some kind of motirsed vehicle based on two servos. This two servo will be fixed to have continious rotation, and will be attached in one end, in the other end i wil sit, my weight on this side will be supported by supporting wheels, im sitting as far as possible away from the servos, to give them as small amount as weight on them as possible.
The estimated weight on this vehicle will be around 80 kilos.

This is still just an idea, no drawings or calculations has been made yet.
So my questions are:
-Will servos or motors be best for this job?
-Im not as interested in max speed, as long as it can move, but how much torque you think the servos will need for this job? (I know theres a calculator for this here somewhere, but i cant find it)
-If the servos are bigger, and with more power/torque does also the electronic stuff, like microcontrollers have to be bigger (more expensive) as well?

I tought of this: http://www.trossenrobotics.com/store/p/5106-Sabertooth-dual-10A-motor-driver-for-R-C.aspx controller, and a pair of 12 volt motors.
You think theese will be to weak to drive me?

[airman00]

use wheelchair motors , servos that big are 600 bux or more \

If you have more questions please start a new topic .

[paulstreats]

In case you are all wondering I would be using two wheelchair motors and four windshield motors. Wheelchair motors for hips and the other motors for the other joints.

Dont forget that if your 2 larger and therefore heavier motors are at the top, then the smaller motors lower down would have more of a job to support their weight. (Say your robot lifts its right leg up to take a step. The motor lowest down on the left leg has to provide torque to support the other 3 windshield motors and 2 wheelchair motors).

Maybe there is a way to increase the size of the limbs simply by doubling up on the servo's. So 1 at either side of the joint, theres just the problem of correctly synching there movements.

[airman00]

Thanks for all the replies so far guys , its really been a great help

paulstreats you are 100% rite. I also think that the wind shield wiper motors are waayyyy too weak to do this job so I might have to end up doing all wheelchair motors. As for doubling up , I think that will end up being necessary certain joints.

Wow the more I research and design this thing , the more I realize how beastly gigantic and dangerous this thing will be

[airman00]

o btw I ordered the BRAT kit and I plan to do some experiments to see the maximum weight that biped can hold before it starts walking like a drunken robot. Then I can see the relation between robot "soberness"(walking balanced) , weight on robot, and torque of joints.

[Ro-Bot-X]

I hope you looked at this http://www.xrobots.co.uk/android10.htm like Admin showed in an older post. The guy used electric screwdrivers to move the pull/pull tendons made out of string and balanced by ruberbands.

For your ideea of using electric windows motors (the windshield wiper motors may have a built-in reversing mechanism), try to build a controller board for each of them, having a potentiometer monted directly to the joint closing the feedback loop. It's like building a bigger, custom servo. Kind of like this one:



More info about the monster servo in the image here: http://www.lynxmotion.com/images/html/ht01.htm

I don't think you may be able to use a servo board for a big motor, or controll a powerfull motor driver with a servo board, since the servo board takes the back EMF from the motors to adjust the controll loop (and the motor driver does not provide that).

Edit: Actually, Lynxmotion sells RC controlled single motor driver, 6-16V and 25 amps continuous (at 13.8V). All you need is the pot for position feedback. Here is the link: http://www.lynxmotion.com/Product.aspx?productID=493&CategoryID=10

[airman00]

thank you very much RobotX - that motor controller looks pretty affordable . Rigging it so that it has a servo interface is a great idea.