Robot tracking robot

[rehon87]

For a project I am making a PacMan sort of game between 2 robots. However, I need a way for the chasing robot to track the other. I was reading and so far saw GPS and RF signal tracking...but I curious of how accurate these are. I really need something to become inches between the escaping car. There will be obstacles in the course for the cars. Any suggestions would be great!


Best method? Best accuracy?

 

[dknguyen]

GPS is not accurate enough. RF is hard to get directional enough and to follow something you need directionality.

Inches is a hard specification...how fast are these things moving? SLower is definately easier. I would try IR (think heat-seeking missile). Of course, you also need something extra to prevent the chaser from bumping into the leader- but that's easy.

 

[rehon87]

Well as far as speed goes...the plan is the person will drive an RC car with some sort of transmitter. But we will tone down speed to whatever is needed to work. I would like to get it as fast as possible though.

On working with the IR idea for tracking...does the fact that obstacles in the course cause a problem? I plan on using IR LEDs+photodiodes to do the obstacle avoidance. Or perhaps I would just need to raise the tracking IR signals above the obstacle heights?

 

[Mikebits]

This sounds a bit difficult. Perhaps you should try something simpler. Is this for a class project?

 

[rehon87]

Well its a EE senior design project and I have a lot of time to complete it. I will be programming a MCU to control the robot and avoiding obstacles, and I can do all that just fine, as I have done something similiar before. I am just trying to find some good ideas for telling the chasing robot where the RC car is.

 

[Mikebits]

I think DK is the resident expert on robots.

 

[Mikebits]

What if you made tracked robot tell tracking robot when it is getting closer? Perhaps a vertical asymptote function. Each robot could have a T/R device and relay information.

 

[Mikebits]

One more thing. Having Car A controlled by a person will make things way more difficult. Perhaps Robot A follows a dark black line. Robot B just tracks robot A.

 

[duffy]

Hey! I designed and built one of these several years ago. Used a ring of IR sensors on the pursuer and IR led's on the pursuee. Worked great, but there are reflection problems from table legs and mopboards if you run it indoors - so your obstacles need something that doesn't reflect IR well. I found brown wrapping paper and black construction paper work well. Get an IR "night vision" scope to check for reflection problems.

You need both frequency detection (for long distance) on the receiver and A/D conversion on the level of the IR recieve signal (so it knows when it's too close). You differentiate the recievers so you know what angle the "cat" is relative to the "mouse". I didn't use GPS or any of that.

 

[dknguyen]

RC car speeds is really fast. It's definately easier at "slow walking speeds" like 30cm/sec. RC car speeds would basically be building missiles lol except you have the added challenge that the chaser can't actually impact the leader. I'm going to guess this is only for horizontal surfaces right? With little elevations and hills then one vehicle might lose track of the other unless it had something like a pan-tilt sensor head to allow the vehicle to look around without actually moving the vehicle (or just make the sensors have a really wide field of view vertically and horizontally.

Getting it to follow at a distance of a few inches at speeds of 50km/h has a bunch of problems. A big one being that even if you had something that could follow that closely, it would never be able to stop in time to not collide with the leader- somewhat of a problem since any maneuver would make the leader slow down (ever heard of tailgating?). Slowing things down is the best thing you can do to make this easier.

The faster you go, the larger spacing between the two vehicles you need. I can see why you want inch spacing though because you are following around obstacles and you need the distance close enough so the chaser can blindly follow the leader and not need to "think" about how to catch up with the leader.

Personally, I think something vision-esque is needed (with IR being the simplest). Heatseeking works too but those are harder to pull off and easier to fool. You could get a CMUCam or some other hobbiest vision camera system and apply an IR filter to it and get use the onboard processor to calculate the centroid of the image which will probably be the IR beacon on the lead vehicle. CMUCams are really cheap for hobbiests, but if you are on a school project budget then they might take up 50% of your budget.

So in that case, a manual array of IR detectors is what you have to use (like Duffy).