New Humanoid Controller Design Input

[lampcov]

Hey Everyone,

I have been presented with the unique opportunity to provide input to a company looking to offer a humanoid robot kit. They currently have a controller that uses the Atmel 89C52 microcontroller, and are looking for suggestions on what their next generation board should use.

I have seen the popularity of the SSC-32 and think the ATMega8 or possibly ATMega128 used for the RN1 could be good choices. Not that they want to steal thunder from the SSC32 (Lynxmotion is our friend).

If you had the opportunity to have input on what YOU would want in a controller for a 17+ DOF humanoid what would you suggest. I'm new in the area and thought I'd reach out for some opinions. Obviously we want to keep the cost down, so built in WiFi isn't gonna happen, but I would appreciate any "realistic" suggestions and input. Who knows, if this works out, maybe we'll have another bot for the family...one built with our desires in mind.

Thanks,
Marc

[Robo1]

I would go for the atmel168 as it's small but with plenty of power 20 MHz. If you want a wish list a 1 million gate FPGA would be a dream.

what servo's are there going to use PWM or digi.

Bren

[lampcov]

Bren, thanks for the input, anything else will be appreciated.

The servos will be controlled via PWM, here are some general specs:

Double-roller, steel gears, copper output Shaft
Voltage : 4.8V ~ 7.4V
Torque : 14kg cm
Speed : 0.2 seconds / 60
Weight : 56g
Maximum action angle : 185

The controller should be programmable by C, C++ and the company has committed to building a GUI interface similar to those used by the RN1, Bioloid, and new HPI/Futaba G-Robot. I am thinking that it will be bundled with a PS2 type controller for the casual user, have a GUI interface for making your own motions (the hobbyist), and will use C/C++ for the "Hacker".

The goal is to offer a robot that is cost effective to buy/run(more so than what's available now) and provide functionality to the Casual user, Hobbyist, and hacker alike. That is why I'm looking for the communities input. I'm new to robotics, but not to marketing/product management, and I see a great opportunity to bring a new product to market the way WE would want it.

Again, thanks for your input and keep it coming! Feel free to e-mail me direct at lampcov@botconnect.com

Marc

[Robo1]

Have you thought about using something that is already built like the gumstix that would give it some real processing power. also using openservo.com boards for the servo so that there digital.

bren

[NovaOne]

I agree with Bren, Gumstix type modules running Linux seem to be the logical next step.

However the step may be too large for your company, and not cost effective?

But I dont see why you should stick to 8bit controllers, when there are so many good 16 and 32bit ones around. I think these would make doing the dynamic walking calculations with IMU's a more viable option, meaning your company will not be offering same-old same-old or worse a obsolete product.

I would personally like to see a servo controller board with that utilised several Parallax Propellers?

[lampcov]

Bren/Nova-

Thanks for the input. I really appreciate it. I would love to use something like a gumstix, I think that would be great, and who knows, if things work out maybe we can bring a "High end" product to bear with something like that.

For this initial offering though I don't think that will be an option, unfortunate as that is. So assuming we went with something like at ATMega128, what features would you like to see on the controller?

Marc

[NovaOne]

Ok, heres my list:
More than one UART to comm with additional controllers as required.
USB for PC connection.
High level support for SPI and I2C.
Modular design for expansion.
Specific ports and high level support for SparkFun -IMU 5.

Fast digital comms with servos (openservo.com board compatibility).
PWM's days are numbered?

[lampcov]

Well I've got good news, and bad news, and need a little help,

We have a nice new compact controller. It uses an ATMega128, has 1MB onboard memory, tri-color LED, 6 Octave sounder, runs 24 servos, 4 Analog and 8 Digital I/O ports, IR Receiver, 2 programmable pushbuttons for program execution, and RS232 connectivity.

It's a nice board and while not a gumstix, for our entry into the market we feel it will be a good place to start. That's the good news.

The bad news is, our lead (read:sole) AVR developer has had to leave due to health problems. He was unable to complete the firmware and we are in a bit of a spot.

I have seen a number of people around here who were looking to have the chance to make their own controller. So, the question goes out to the group...

Are there any interested parties in taking on the task of finishing our existing code, or if you have a better plan (which could be) implementing that. I'm the marketing/sales/CEO guy, not a dev, so I'm not much help from a technical standpoint, but I should have all the documentation for the board and existing code soon (it's in Chinese but translation is coming). This isn't a charity and we are willing to contract with someone to work on this.

If you are interested, please either PM or email me directly at development [at] robofactors [dot] com.

Thanks,
Marc

[limor]

I hope you know what you are doing.
there are at least 3 new hobby servo humanoid coming up from Asia at under 400$.

[trey3670]

care to elaborate on those three?