Let's Make Robots!

Robotic Ants Farm

Well, Aniss1001 brought up in an other topic the idea of creating a swarm of robots. This reminded me of an idea I had 5 years ago, when I was in the States and building Lego Robots. Funny thing, a quick search on the net at that time revealed several similar projects. Also, my inexperience at that time with non Lego robots didn't let me put it in practice.

So, here's the idea:

A closed environment, something like a fish tank, but perhaps lower walls. The floor painted black, walls grey. There are 2 opposite corners or sides where a white line delimits the "home perimeter" for the robots. Inside that perimeter there is a charging station with aluminum foil on the ground and ceiling so the robot's feelers can touch both to charge the batteries. The charging power comes from solar panels, so the entire project is self sustained. The charging station emits a coded IR beam that all robots can see so they can find it when charge is needed. This is mounted on a pole like a flag. The robots are divided in 2 teams, each team has the same (power LED) color (red or green). Each team "speaks" different IR frequency (something like 32kHz and 56kHz) so they don't interfere too much. Besides the 4 IR receivers (and emitter), the robots have bumpers, an IR eye and a mandible, so they can detect and grab small white foam cubes that are spread around in the environment. People can add more cubes from time to time through a whole in the environment ceiling, like feeding the fish. The robots search for the cubes and bring them home to build a wall to protect their home and flag. They will also try to capture the other team's flag and bring it to their home. Inside the home perimeter, a "queen" robot is using the cubes to actually build a 2 layer wall.

Until the HXT900 servos became available (so cheap) I had no idea what kind of motors I could use for the robots, so I can build them small enough. Now, I think this project can be done nicely with a budget of about $200 for 10 robots. One queen and 4 ants in each team. Each robot can be powered by a small cell phone battery and there are solar chargers for them not too expensive.

Some more thoughts:

Robots can be made identical, they could use a bi-color LED and IR receivers for both frequencies so they can switch sides for instance, the queen can "hypnotize" them to get them on her side for a certain time. Or they can pretend to be on the other side to steal the flag...

Anyway, let me know what you guys think! 

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.

After seeing this swarm, I've decided to follow their example and actually build this Ant Farm. But there will be a few differences. They use PIC, I'll use Arduino (a SMD ATmega168). Also, I will use a Maxim chip for charging the LiPo battery. And I will use those IR demodulators I was talking about in my first post. Not sure yet about the Queen, but all the other bots will have surplus pager motors. I'll order the boards from Seeed Studio, so anyone can order them from there and build their own swarm. Everything will be OS. I'll work on a preliminary schematic and make a test bot to see how it performs. Perhaps two bots, to try out the comm, but the code was already done here. I just need to translate it to Arduino.


Thanks for sharing! That project seems really nice.

I was thinking about making a compound short range IR sensor for detecting the cubes and feelers on the mandibles to be able to tell when a cube is grasped. This compound sensor would be made from a few IR LEDs that would flicker at a low frequency and low intensity and 3 IR detectors to be able to center the robot on the cube. The emitted light would be just enough to see the cubes at a short distance so it does not interfere too much with the other robots. The robots can also use this sensor for robot to robot communications. At a higher level, on the top of the robot, there would be another (circular) IR sensor that would detect the higher intensity and higher frequency IR light emitted by the charging station.

As for building the wall, I had a Lego robot prototype that was building walls using Lego cubes along a line the robot was following. Unfortunately I don't have pictures of that prototype, but it worked pretty well. So it is possible. And that would be the queen's job to do. The ants just bring the cubes to the base.

Your idea is nice, but undetectable by the human observer without an explanation of what is going on. The idea of the project is to observe how the robots behave and develop special case algorithms so they can get unstuck or even help each other. And of course, be a nice spectacle to watch for hours...

I've seen quite a few variations on this theme. I like this one - http://people.bath.ac.uk/ensab/B_man/tr_11_00/#100

 However, I'm still not sure if it's practical yet - 

How are the ants going to find the cubes? A white cube would be pretty easy to pick out from a webcam image, but then you either have to mount a webcam on each ant, or use one looking down on the colony, which a PC scans for cubes and tells the ants where to go, which kind of spoils to dynamics of the system as the ants wouldn't be using their own senses.

How are the ants going to build the walls? It's going to be pretty hard to get them to do much more than dump the cubes near the IR beacon, which will clog up the charging station.

Maybe a way to avoid this would be to have them forage for some thing else. Eg. have the queen at one end, and a number of charging stations scattered around, with some obstacles. The ants move randomly until they find a charging station. Then they charge up, and find their way back to the queen, and 'feed' her with some of the energy they've picked up. If you had some way of simulating the ants' pheremones, they should be able to find the fastest route from the charging stations to the queen. Maybe switch charging stations on and off from time to time, to keep them on their toes. 



Sounds similar to the MIT Ants put together and studied by James McLurkin.
I actually read about that some time ago but had forgotten about it. Thanks for providing the link. It has now been bookmarked :)

Ants sure are interesting creatures indeed. When I was a kid I could sometimes sit and watch them for hours. The way the work together in a very efficient system etc. Sometimes I would place minor obstacles in one of their paths and watch how they would get confused and then work together to remove so they could get back to business ASAP.

I also read many years ago about how they communicate using feromones. For instance I read about an experiment where some scientists had managed to isolate the feromone an ant produces when it dies. The funny thing was that when they sprayed some of this feromone on a living ant the other ants would consider it dead and remove it from their workspace. But the "dead" ant would then run back to work as soon as it was put down, thus this could go on for a while.

The idea I had about using a PC as a hive brain and using LEDs as a rudimentary way of communicating came to me because I was thinking about how I could simulate this feromone based communication system in the cheapest possible way. For instance ants use feromones to mark a specific spots indicating different things. Eg. there is food here or not, this path blocked or not, efficient or not (not sure how it actually works...have some reading to do here :)

Anyway I love the very concept of swarm bots. And as you say the current prices of tiny servos (check out DAGU's microservos: 2.40$) and other parts make it possible to carry out. So it is one of the things I'll be experimenting with during the years to come for sure.

I'm thinking each swarm bot would consist of something like: an Atmega168 (perhaps AtTiny), 2 modified microservos, 2 bumpers, 2 LEDs (as mentioned in the other thread), an IR receiver and perhaps a home made IR sensor. It could off course be cool if they could somehow move objects. I was thinking perhaps make some small eletromagnets so they could pick up tiny metal objects. It is off course vital that each robot will be as cheap as possible (hopefully <15$).

Anyway thanks for sharing your idea. I love it and hope you'll carry it out ;)

PS: Also check out ezekiel181's swarm idea if you haven't allready.

PPS: LMR needs more swarm bots :D

Thanks! I know about hobbycity, I got all the servos for my MiniEric robot from them. That's why I said now there are cheap motors available. I would get 3 of them, mod 2 for continuous rotation and use the third for the mandible. A simple mechanism can be made to grab and lift with a single servo move. Paper clips are good enough for that. PCB can be made up to 5x5 cm and ordered cheap from SeeedStudio ($30 for 10 boards). A Tiny84 or 861 should be enough to get things done.

This would be a good candidate: http://www.hobbycity.com/hobbycity/store/uh_viewItem.asp?idProduct=8995&Product_Name=Turnigy_TG9_9g_/___1.6kg_/___0.12sec_Micro_Servo at $2.99 x 30 makes $89.70 plus shipping...

I'd love to see your "grab and lift with a single servo move". How does that work?
Dude this sounds like an awesome idea!! I want to see it done! This would also be something cool to have like in an office cubicle desk or something. Simply awesome. I'm not sure where you wanted to get your servos or how much they cost but http://www.hobbycity.com/hobbycity/store/uh_listCategoriesAndProducts.asp?catname=Micro+Servo+0%2D10g&idCategory=287&ParentCat=189 offers some pretty cheap and small servos.