Autonomous Robot Query

[neo01124]

Guys, I'm working on an Automated Defence Robot for the Techx Challenge (http://www.dsta.gov.sg/techxchallenge) and I'm confused as to what sort of processing unit arrangement i should be using i.e.
1) Use two micro laptops(Vaio UX 380N) one for Image processing and other one for processing rest of the sensor data (LMS,GPS,Accelerometer,IMU,etc).
2) Instead of using two laptops  customize  a micro motherboard (Mini ITX) for image processing and use a really fast microcontroller for rest of the processing i.e. a dsPIC.
I am biased towards the first method since there will be really less hardcore electronics since i'm not really good at it.I need u experts to tell me the pros and cons of the above two methods helping me to evaluate the best possible one.Any other suggestions will be welcome.Thanks in advance!!

[neo01124]

15 views and no replies whats the matter guys??

[JonHylands]

Get a dual-core (or quad-core) mini-itx...

- Jon

[Ro-Bot-X]

I would go with the second ideea. Interfacing sensors directly to a motherboard (laptop or other) is not easy. Simple touch sensors and relays may be interfaced, but analog sensors require analog to digital conversion that a motherboard doesn't have. To go the easier way, get a serial or USB sensor interface and a USB servo or motor interface and connect them to the Micro ITX motherboard of your choice. In this way you don't have to worry about microcontroller programming. Your software will have to be layered, lower layer monitoring sensors and send commands to actuators, higher layer doing vision and A.I.
Does this make sense?

[neo01124]

"To go the easier way, get a serial or USB sensor interface and a USB servo or motor interface and connect them to the Micro ITX motherboard of your choice.:"
How does this eliminate the need for a analog to digital conversion, i mean do the usb sensor interfaces have a built in adc?? Which usb sensor interfaces are the standard ones used in robotics??
"Your software will have to be layered, lower layer monitoring sensors and send commands to actuators, higher layer doing vision and A.I."
Please throw some light on this part of the discussion.

Whats wrong with the first approach according to you?? It would eliminate most of the hardcore electronics and we could concentrate mor on the coding part.
Thanks

[Admin]

You cant just directly attach motors and other hardware directly to your laptop - so you cant avoid the electronics.

You should decide on your requirements before deciding on the hardware. If you can determine how much processing power you need, you can determine which hardware choice is best. Tell us more about what algorithms/sensors you plan to run/use.

And don't forget the palmpilot option

[Ro-Bot-X]

Yes, usb interface boards have microcontrollers on them that are allready programmed so you can concentrate on the PC programming part. Check this web site, it has all the usb interface boards you may need, including sensors:

http://www.trossenrobotics.com/phidgets.aspx?a=pg

Use USB hubs to connect more devices to a motherboard.

Layered programs are programs that run in parallel, sharing common global variables. They are layered because they have different priority at sharing processor time. There are programs that work with the hardware, like a network card driver (low level layer), and programs that work with artificial intelligence (high level layer). They are exchanging information so they can work together like there were just one program running.

Does this makes sense?

[neo01124]

You cant just directly attach motors and other hardware directly to your laptop - so you cant avoid the electronics.

But we would be minimising a whole lot of it with this arrangement.

You should decide on your requirements before deciding on the hardware. If you can determine how much processing power you need, you can determine which hardware choice is best. Tell us more about what algorithms/sensors you plan to run/use.

We are goin to use  Sick LMS for obstacle detection, a firewire camera( i dont remembr xact name rght now) for higher level vision, a GPS receiver for navigation and a 3 axis acclerometer for feedback for critical situations like stairclimbing,etal.Moreover its goin to run some complex code for control of an actuator. I dont really kno a way to determine the processing power requirement for these, how is it done ??

Layered programs are programs that run in parallel, sharing common global variables. They are layered because they have different priority at sharing processor time. There are programs that work with the hardware, like a network card driver (low level layer), and programs that work with artificial intelligence (high level layer). They are exchanging information so they can work together like there were just one program running.

Does this makes sense?

Thanks it surely makes a lot more sense now.What sort of an OS should be used on such a system(Laptop one)?? My study tells me Linux .Does it support 'C' and can this(Layered programming) be done in 'C'?
Thanks for all the help guys!!

[Admin]

We are goin to use  Sick LMS for obstacle detection, a firewire camera( i dont remembr xact name rght now) for higher level vision, a GPS receiver for navigation and a 3 axis acclerometer for feedback

Whoa . . . first, you are asking for waaaaay toooooo much for your first robot. 'high level' camera vision, SICK sensors, GPS . . . You need to back off a little and stick with the basics first . . . Not to mention the cost of all this . . .

Every level of complexity you add is another ~month on development, and an additional layer of problems you will encounter.

That being said . . .

So SICK sensors have a very high data rate, and the processing required depends entirely on your algorithm. Look up research papers on the algorithm you plan to use for an estimate on required processing. Same for the camera. GPS requires a serial connection, and the accelerometer (what, no gyros?) will require ADC (which requires a microcontroller).

for critical situations like stairclimbing,etal

Just know that if you use a laptop, the harddrive can break from vibration and collisions. Thats the advantage of microcontrollers - they are solid state so can take a severe beating without breaking. Not to mention the huge power savings . . .

complex code for control of an actuator

actuators don't really require complex code . . . which actuator?

I dont really kno a way to determine the processing power requirement for these, how is it done ??

Datasheets and research papers. You just gotta look them up.

[JonHylands]

If you can't figure out the answers to these basic questions you're asking, there's no way you're going to be able to do anything with the advanced sensors you're talking about...

You're talking huge amounts of research here.

- Jon

[neo01124]

Whoa . . . first, you are asking for waaaaay toooooo much for your first robot. 'high level' camera vision, SICK sensors, GPS . . . You need to back off a little and stick with the basics first . . . Not to mention the cost of all this . . .

The cost wont be a problem cause we got some corporate sponsors backing us but i kno its a bit over the top but i'm and will keep on working hard as hell to get there!!

actuators don't really require complex code . . . which actuator?

I'm talkin abt a 3dof arm like thing which'd be pressing elevator buttons.

Datasheets and research papers. You just gotta look them up.

I have looked up a lot of research papers and project reports and have an idea of the processing demands but i wanted to kno whether there was another more accurate way (system specific) of gettin to the answer.

You're talking huge amounts of research here.
- Jon

No offense Jon but i never said i was wary of hard work but i just wanted to get an xpert opinion so that i dont waste my time with trivial useless stuff.
That said Thanks for your replies!!

[Admin]

We dont mean to offend you but . . .

What me and JonHylands meant was that your project is too ambitious and will fail for that reason.

My first robot was a 1/10th as ambitious as yours, and it failed. It failed because I tried to do too much, knew too little, and the mistakes overwhelmed me. It has nothing to do with you being a hardworker . . . even an expert would struggle with your project. As a matter of fact, its beyond even state of the art - no one ever has accomplished this.

I can still guide you but honestly, you should first understand how a simple photovore can be built . . .

[neo01124]

i kno its ambitious but i never said this wud be my first robot i am going to work on a series of small bots just like everyone else does and keep on workin on this one all the while and achieve the whole of this over a period of time
neways i'll keep posting my doubts over here and wait for ur xpert opinions if thats allright

[neo01124]

which books or text shud i be readin to get my head into AI??
it'd be cool if someone cud post links for ebooks etc
thanks in advance

[Admin]

Society of Mind is my fav AI book.

Others have recommended On Intelligence