Autonomous Sentry Ideas

[3DWolf]

So I've got some ideas cookin in my head on how to recreate the Level 1 Sentry from Team Fortress 2 (airsoft version; non-lethal, don't worry), but I've come to a slight roadblock - the microprocessor.

I need one that's small enough to not destroy my wallet, but also powerful enough to be able to either use motion detector data or parse a camera signal and be able to read objects from it. I've seen some other platforms that use the latter method but they used a laptop next to the unit to do all the processing, but I want my unit to be enclosed within itself (aka no laptop).

Planning on using servos to move it, not quite sure which of those I'm going to need to use yet or not. Depending on the weight of the system I might have to use a high-torque one which I really hope I don't have to shell out the money for.

Anyone have any idea's or know of any good microprocessors that I can use for this? Preferably that support C or C++

[PAR_WIG1350]

It isn't technically a microprocessor, but the Roboard RB-100 is a full computer for around the cost of a NXT kit (250 alone, 300 with everything). It has 24 pwm/servo io channels that can be regonfigured in software as gpio. It also has 3 usb ports. It isn't very well equipped as far as memory, but the CPU runs at 1GHz. http://www.robotshop.com/roboard-starter-kit-3.html

I'm not sure giving anything autonomous the ability to fire even airsoft pellets is the best idea -->, although it would be cool for laser tag.

[RyanCahoon]

You have a couple different choices of sensing technologies; which one you pick will determine what power you'll need in your processor board.

If you want to go with full customizable vision, you'll need something on the level of the RoBaord that [PAR_WIG1350] pointed out. I would personally point you toward the Surveyor Blackfin board ($195), since it already has the camera integrated. It's programmable in C; I found it pretty easy to learn the API they've created. They've also come out with Linux support recently if that's easier for you. You can get it with Wifi for $350 if you want to have it report back to the mothership...

For slightly reduced functionality, you might want to check to see if your team has any of the CMUCam2 board from 2005/2006/2007. We only used the color blob tracking function for FIRST, but they also have a simple motion detection/frame differentiating function that can be run. You could pair one of these with a low-powered microcontroller (see below) that would control the servos.

Taking a different approach, you might could use some sort of proximity sensor. Options include PIR sensors (what they use in automatic doors), or ultrasonic or infrared rangers. You should be able to get any of these for less than $25. If you only want a proximity alarm, you should only need one-three (depending on the desired monitoring angle and sensor type). You may have to go to a higher number if you want to actually aim accurately. You'll have to be careful if you attempt this with the range finders, as they rely on sensing reflections, and crosstalk between the sensors can invalidate the readings.

These types of sensors would only require a fairly low-power processor. The common choice would be an Arduino (as is the case in the linked PIR sensor tutorial), but if you want to go really low cost, TI just came out with a new controller board that costs less than $5 that looks interesting.

--Ryan

[JamesBrown]

Originally Posted by RyanCahoon

Taking a different approach, you might could use some sort of proximity sensor. Options include PIR sensors (what they use in automatic doors), or ultrasonic or infrared rangers. You should be able to get any of these for less than $25. If you only want a proximity alarm, you should only need one-three (depending on the desired monitoring angle and sensor type). You may have to go to a higher number if you want to actually aim accurately. You'll have to be careful if you attempt this with the range finders, as they rely on sensing reflections, and crosstalk between the sensors can invalidate the readings.

These types of sensors would only require a fairly low-power processor. The common choice would be an Arduino (as is the case in the linked PIR sensor tutorial), but if you want to go really low cost, TI just came out with a new controller board that costs less than $5 that looks interesting.

--Ryan

I have seen a Sentry that works using PIR and an Arduino, it is accurate and pretty cost effective (the Arduino is overkill for this application but with all of the resources the community provides I would still recommend it)