RoboSavvy Forum

Newb

2005-03-17T15:32:44+01:00

inaki wrote:
Derek, might you ellaborate on how you think to use Gumstix in your project ? You know people here (well, mostly Limor ) is interested in this platform too.

I'm not exactly sure yet, it's just such a cool and tiny board that I want to use it!

There are gumstix boards with built in bluetooth, so I could use that to replace the radio scheme I am using now, which uses a rather large 8-channel 75mhz receiver.

I also am toying with some ideas for controlling the robot that will benefit from a lot of processing power, and the gumstix can provide that.

Statistics: Posted by DerekZahn — Thu Mar 17, 2005 3:32 pm

Newb

2005-03-17T14:56:01+01:00

Derek, might you ellaborate on how you think to use Gumstix in your project ? You know people here (well, mostly Limor ) is interested in this platform too.

Statistics: Posted by inaki — Thu Mar 17, 2005 2:56 pm

Newb

2005-03-16T23:47:58+01:00

limor wrote:
here's another link to a western built humanoid, using "C555 microcontroller and PicoBytes PicoPic servo controller" though i think the picture is of a gumstix board http://happyrobots.com/bing.htm

Hi! I just discovered this site. I'm the builder of Bing. I will get some more detailed information on my web site pretty soon; right now I'm working desperately to get Bing ready for RoboGames in San Francisco next week. If it successfully walks I'll be happy.

The current version of Bing does not use a gumstix board, but the next version will use one (or at least I think it will!).

I have learned a lot getting Bing as far along as it is so far, but am looking forward to correcting its design flaws in the next version.

Statistics: Posted by DerekZahn — Wed Mar 16, 2005 11:47 pm

Newb

2005-03-08T10:45:12+01:00

I think the gumstix issue deserves a thread of its own..

Statistics: Posted by limor — Tue Mar 08, 2005 10:45 am

Newb

2005-03-07T13:59:33+01:00

Gumstix uses a general purpose processor, the XScale by Intel. That one is an ARM compatible CPU, one of the so called 'System on Chip' processors, because the processor integrates a lot of peripherals. The good side of such a processor is that is is powerful enough to make a complete system. The bad side is that you must use a full fledged operating system to run it. That is to say, you cannot pretend to start this processor and start doing input/output to a digital pin for example, like you would do on a microcontroller. The Gumstix comes with a version of Linux. So it means everything you do must been done through this operating system. There are other operating system available in case you don't like Linux but then you should count on adapting the operating system to the Gumstix board which may be not a trivial task. This may be good or bad news for you, depending on your preferences and your ultimate goal.

For my taste sys on chips are somewhat overkill for domestic robots. Sys on chip processors may be interesting if you pretend to have a real computer on board rather than a controller board that is most suitable for microcontrollers.

Basic Stamps work very well as peripheral controllers. The problem with Basic Stamps is how to do many things at the same time. They are good for one or two tasks but no more. However you can add several Basic Stamps together to perform complex tasks. The best of BS is that are simple to program and have many peripherals already made for it.

For general purpose control I prefer PICs or AVRs. There are many microcontrollers out there, some of them very powerful, but having a common controller like a PIC or an AVR gives you more comfort when you have to do the real development: programing tools, libraries, thousand of available projects already made, documentation, ...
PICs and AVR are inexpensive for the power they give and you can use several on the same board.

For servos control the best choice is a separate controller that can be controller serially instead of relying on the processor direct control. A typical robot has 12 or more servos which is too much for some microcontrollers. Also note that microcontrollers share lines so if you reserve many pins for one task you lose lines for other tasks. There are dozens of them that you can adapt to your board of choice. For initial programming it is best to have a serial servo controller that you can plug to your PC.

There are GPS modules available for robots but the problem with these is accuracy. A GPS system does not give more than one meter accuracy in the best case which is of doubtful interest in a robot (not counting they do not work indoors).

There are many solutions for Bluetooth, one of the most interesting is the EB500 series made by A7. WiFi however is more expensive to implement. There are several WiFi solutions for robots based on the Prism PCMCIA board but this requires the apropriate interface which is not easy to implement. For a robot I believe Bluetooth is the right choice. It can reach 100 meters at 150Kbps which is more than adequate for most purposes.

Cameras for robots are usually analog. They are small and you don't need any special control for them other that using perhaps an arm to mount the camera in order to change the focus. There are literaly hundred of miniature cameras available that are usable for robots. Most of them require a RF decoder on the receiver side. As the camera does not interact with the brain on board you can chose it based on size, power requirements, color/B/W, etc.

As for sensors, there are thousand of them. The most obvious choice for starters are, for obstacle avoidance, infrared detectors, ultrasonic rangers and mechanical switches; for bipeds, accelometers and gyros; for general orientation, electronic compass and encoders. You can add many more depending on your robot needs: line followers, color detectors, pyro sensor, thermo sensors, etc. Some of them are inexpensive while others are very expensive.

Statistics: Posted by inaki — Mon Mar 07, 2005 1:59 pm

Newb

2005-03-07T01:00:45+01:00

OK here goes it so i have just got done ordering like 200 $$$ for aluminum brackets, lexan, screws mounts and stuff like that the thing is i need a board to run my bot moneys isnt really an obejct but i need some advice on gumstix and the type of software i could run the board i have now is a cheap B.O.E stamp 2 sx board from paralex no servo controllers built in or anything i need a board through some out in the air for me and oppions on them manily i need speed and muti function for things like GPS, wi-fi or blue tooth, camera and multi senors i would do the research but i am out on convoys everynight and worrying from getting myslef killed in iraq so please help ...lol

Statistics: Posted by Genocide — Mon Mar 07, 2005 1:00 am

Newb

2005-03-05T01:40:54+01:00

here's another link to a western built humanoid, using "C555 microcontroller and PicoBytes PicoPic servo controller" though i think the picture is of a gumstix board http://happyrobots.com/bing.htm

Statistics: Posted by limor — Sat Mar 05, 2005 1:40 am

Newb

2005-03-03T05:04:10+01:00

I'm from Holland

Statistics: Posted by Meltdown — Thu Mar 03, 2005 5:04 am

Newb

2005-03-03T00:46:32+01:00

I am in America.

I'l take a look at those 3D design tolls limor, and hopefully learn how to use one.

Statistics: Posted by EC — Thu Mar 03, 2005 12:46 am

Newb

2005-03-02T14:09:04+01:00

My choice is aluminium parts. Lexan cannot be folded which limits the shape of parts. Of course there are tons of small parts that are not the aluminium parts themselves to allow articulation. The KHR-1 is the only decent robot I know that uses the hones in the motor as the only articulations in the robot.

As for the motors, really are a crucial part in any design: whether they have free hones, how many supports, how much torque, etc, etc. Not counting the so called 'smart' servos. The obvious election in Eruope/Amercias would be Futaba or Hitech. By the way Kondo servos are sold out of Japan ?

As for the brain. Wow!, the Gumstix is too much I think. I have used embedded XScale boards for a while but it seems a little overkill for a small robot I think. I would look more on the microcontroller side: perhaps PICs or AVR. The problem with these is lack of enough RAM memory. But in my experience many things can be done without too much RAM. The boards can be created inexpensive and are easier to interface. On the other hand Linux is not a suitable operating system for real time and control.

It would be interesting to know which products are easily available in each country.
By the way. Where is located each one of us ? I am in Spain.

Statistics: Posted by Guest — Wed Mar 02, 2005 2:09 pm

Newb

2005-03-02T12:18:53+01:00

EC wrote:
I like the idea too. But I can't really help out much, due to my lack of money, experience, and education...

Perhaps I could offer moral support?

Sure you can.

learn how to use a 3D design software and try to create humanoid designs..

There are many 3D design systems.
My favourite (not-free) system is 3D Studio (there's a free version of it too) but the one i'd recommend, which is one of the greatest opensource projects ever, ishttp://www.blender3d.com

have a look at some robot designs :
http://gallery.mcneel.com/?language=en&g=12

Seems like the tool of choice for the robo-one crowd is Autodesk Inventor. on the robo-one site there are several links and i found this one which is a tutorial for building humanoid parts : http://www.world-c.jp/robot/inventor/inventor005.html

And here's another thing google digged out; someone from Italy did exactly what we are talking about:
http://www.jodinsky.com/index2.html

Statistics: Posted by limor — Wed Mar 02, 2005 12:18 pm

Newb

2005-03-02T01:25:04+01:00

I like the idea too. But I can't really help out much, due to my lack of money, experience, and education...

Perhaps I could offer moral support?

Statistics: Posted by EC — Wed Mar 02, 2005 1:25 am

Newb

2005-03-02T00:46:13+01:00

Anonymous wrote:
It is interesting to note that we all want to do the same thing, but I should admit it is not easy: build a humanoid/walker robot by ourselves.

Building aluminimum parts is the most difficult part for me. I am skilled at programming and electronics but I consider building mechanical parts the biggest challenge of building any substantial robot.

In the KHR I appreciate the fine building of aluminium parts. However I miss the electronics parts: autonome processing and sensors. Perhaps the KHR has too much power to be controlled by a common microcontroller ? I don't think so. I brain for the KHR would be great and some sensor are really necessary.

Anyone interested in building a new humanoid from sratch ? Perhaps improving significantly the KHR-1 might be an interesting first step ?

I am interested.

Great idea!

Here's a braindump of steps towards implenting this project as i see them..

1) choose the servos to be used
i'm afraid a design is hard-coded with the type of servos used because they come in specific box sizes and shapes, torques etc. (hopefully there's going to be enough participation to make more then one walker design)
2) choose the controler board
there are many controler boards out there and the new ones with the ATmega CPU come with zillions of PWM outputs and loads of 10bit A2D inputs. or we can go for my favourite : Gumstix and its recent PWM addon board
3) design aluminum or lexan parts
this involves obviously a using 3D design (CAD) tool. the biggest challenge here is creating common parts. KHR-1 has an ingenious bracket (see the bottom of this page: More Brackets) that is reused in different parts of its body. the problem with having many parts is the fabrication cost. Lets say we manage to nail down 5 core generic parts. The price of emachineshop.com then goes down dramatically when ordering large numbers of the same item making this robot more affordable.

We could addopt a creative commons license which could attract the likes of SlashDot.com community and in turn allow prices to go down...

Statistics: Posted by limor — Wed Mar 02, 2005 12:46 am

Newb

2005-03-01T20:19:18+01:00

It is interesting to note that we all want to do the same thing, but I should admit it is not easy: build a humanoid/walker robot by ourselves.

Building aluminimum parts is the most difficult part for me. I am skilled at programming and electronics but I consider building mechanical parts the biggest challenge of building any substantial robot.

In the KHR I appreciate the fine building of aluminium parts. However I miss the electronics parts: autonome processing and sensors. Perhaps the KHR has too much power to be controlled by a common microcontroller ? I don't think so. I brain for the KHR would be great and some sensor are really necessary.

Anyone interested in building a new humanoid from sratch ? Perhaps improving significantly the KHR-1 might be an interesting first step ?

I am interested.

Statistics: Posted by Guest — Tue Mar 01, 2005 8:19 pm

Newb

2005-03-01T16:52:41+01:00

yeah that would be cool what all can you do with a CNC machine, and by the way i think that the emachine shop website could be good i would like to have things that have curves and bubbled out, i am tottally new to this stuff and i am waiting around on my KHR-1 with in the next 2 weeks or more where i am in Iraq i have a 10 day wait any other companys that have good deals on biped type robots with the same quality.

Statistics: Posted by Genocide — Tue Mar 01, 2005 4:52 pm