Robot Wall Racers
- Actuators / output devices: 2 motors each
- Control method: autonomous
- CPU: Picaxe 28X
- Operating system: Picaxe
- Power source: 4 CC cells
- Programming language: Picaxe basic
- Sensors / input devices: SRF05
- Target environment: any room
UPDATE: Nothing new, just a small edit. Am planing a real update one day :)
A video in better quality than YouTube & Google is here
How to make Robot Wall Racers yourself, see here
The car-chase was made by holding a camera on a stick, chasing the cars around. Some clips where made by placing the camera on a record-player. All editing was done in the standard "free" program that comes with a Mac, called iMovie. It is a nice program for light editing.. but phew, it was pushed to make this
The second video shows some work with the skid-turning. It also shows improved handling in general.
Notice how it sometimes mis-judges when the wall is not 90 degrees, but has bumps. But also how it handles getting out again :D (Me proud, love this, think it is cool, it drives so much better than what I can do with remote myself)
Of course I could just give it some more space to turn in the code, to drive more "safe", but I kind of like it to be tight :) And notice hw it never bumps into anything else - the polystyrene-blocks would fly right off if it hit them.
I have added an extra battery for more power.
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>Your home is a racing track, you just do not know it.
Look at your floor where it meets a wall.
Between 30 and 70 cm from that goes an imaginary path, following along the wall.
That is the racing track!
Where the wall meets another wall, or there is an obstacle, the track follows in a smooth path, always 30-70 cm away, on the floor.
The track returns to where you started, you see! :)
It is a full racing track:
You can time how long time it takes for a little robo-car to take one round. You can then try to improve the code, and time it again.
If you make the car like this:
One sensor (Could also be infra-red or whatever) looks ahead, another looks to the right. Then code could be something like this:
If something in front, turn left.
If something closer than 30 cm to the left, turn left.
if nothing in front and something closer than 70 cm to the left, drive ahead.
If nothing to the left closer than 70 cm, turn right.
A car is racing your home!
It can be timed how long before passing by the same line again and again.
But a single car is just test-labs, no?
Now take 2 cars with the excact same setup as above, let's say a red car & a blue car. Now it get's interesting!
If red car is behind blue car, what does it do?
Well.. Something (blue car) is in front, so we turn left to take over.. until nothing is on the right, so we pull back.. That is trying to take over!
At a point the blue car is turning left because there is something in front.
Then the red car will have the inner track..
What is actually going on, is that one car is behind, but is "strugeling" to take over.. when it is in front, the other will fight to get ahead.. and we have a full race :D
You can take Red & Blue and place them next to each other, next to the wall; They will both drive full speed ahead, until a turn, where the inner car will have advantage. But then..
Wall Racers are born, ladies and gentlemen.. I had these 2 cheap RC cars with crap maneuvering abilities, and thought of what I could do with them. Crap - but still; They sort of drive like cars - not like differential steered robots.
It has ben quite simple, have experimented with AZ8222's instead of motor drivers. And just a Picaxe on a standard board, simple and fast.
Of course, when turning angle is not enough to avoid obstacles, maneuvers with reverse and 3-point turns must be done. But this is why we have a microcontroller, and it is not spoiling the fun - on the contrary :)
Good idea! Maybe an IR
thats what i was thinking
thats what i was thinking about while was reading.. =D
go on.. i wanna see the race =P
(can't w8 for my ultrasounds arrive.. i don't have any distance sensor.. only can make it with bumpers switchs or IR reflection.. that bouth sucks..)
No guys - you don't get the
No guys - you don't get the beauty in the math :)
If any extra, all we need is a bumper at the back to ensure it knows if it is reversing into something when doing a 3-point turn.
So far it is just theory - I had the idea late last night. But I might get the time to build them later today if all goes well - so we will see if it works as I hope :)
/ Fritsl
Sounds like fun!
hey frits you know what! you
Not only that - I am going
Not only that - I am going to win it! =) I think!
:D
/ Fritsl
"Toy" That's exactly what it is
Hehehe. What do you think you're going to win with that tiny little thing? THIS will (eventually) do the same thing, but outside in the big bad world!
God created the integers; all else is the work of man - Leopold Kronecker
"The same thing"? My "Wall
"The same thing"? My "Wall Racing-concept"?.. or "just drive around and avoid obstacles"?
/ Fritsl
Haha!
No. I was having a laugh. I've not considered an entry for this competition, although I probably should. I could use those tracks.
God created the integers; all else is the work of man - Leopold Kronecker
Looks like fun. Too bad
CrisTheCarpenter did a
CrisTheCarpenter did a similar project, I remember :)
/ Fritsl
R/C
I've had a R/C remote send an input to a PIC. It's dead easy. The waveform output by the R/C receiver is IDENTICAL (surprise, surprise) to that required as the input to R/C servos which is so well dcouemnted all over the web.
God created the integers; all else is the work of man - Leopold Kronecker
nice pictures =D greath
Oh man, these robots, like
Oh man, these robots, like all of your robots, look so fun. One MORE idea I want to play with :) The speed, I think, is the big think that makes these appealing. My attempt at controlling an R/C car with a microcontroller didn't work very well, because it couldn't process the input from the sensor fast enough to react in time. But these can aparently react even when driving at full speed, and that makes them really fun :)
Dan
Photos and Controllers
It helps that Frits appears to be bloody brilliant at macro photography!!
My experience the first time I tried this (with my car above) was that I had no way of reversing a 10A 7.2V motor! I couldn't find a suitable H-bridge and I was (as ever) on a budget that didn't allow an electronic speed controller. I should have stuck with the mechanical controller and driver the servo. In the end I used a DPDT relay but I didn't understand the concept of waiting for the motor to stop before reversing direction. Yes, there were sparks.
I am using 4 DPDT relays as
I am using 4 DPDT relays as well. TheCowGod showed e that they could be used to revert current (thanks), and I use another to turn the motors on/off.
I Know that 2 of the relays has one set too many pins etc, but yadayada. It was what I had :D See pictures here: http://picasaweb.google.com/fritslyneborg/WallRacers
To both of you; I think the challenge is the programming. My initial attempt did bang and crash a lot, SRF05's where torn off etc.. So I took my own medicine, just started with driving ahead and stopping before a wall, then.. and then..
I would never be able to "just program this". One thing, example; I discovered that it takes a 20 miliseconds (20 ms!!) yank to one side JUST before reversing, and then turning etc to make a skid turn come out right (skid to the right side).
My first code did not have that kind of details, but then I programmed layer upon layer of things that works..
It also helped me to insert a potentiometer before the rear motor, and film the behavior - all to analyze / get slow motion.. Because there is no way in hell I would be able to figure out how to create the code else. These things just fly around, gives a large impact on a wall (they are heavy now), continues to flip out afterwards .. and it is impossible to remember "what went wrong".
But with these tricks it was quite easy, spend perhaps 2-6 hours of programming & testing to get it completely right, and it was easy & doing well from first (simple) line: Stop before wall.
This approach also makes it drive "like a person", IMHO.
/ Fritsl
Rgarding the photos
Regarding the photos, I tried to give my knowledge here:
http://letsmakerobots.com/node/466
/ Fritsl
Coding Nuts
Wow! My code is only 210
Wow! My code is only 210 bytes :D
Am optimizing, will make walkthrough!
/ Fritsl
Awesome, I love the idea,
very very nice,good stuff
very very nice,
good stuff always rise from simple ideas.
hello
I was impressed with your racers, made me want to do my own now :) i was thinking about how close you could get them to the wall,have you tryed putting proportoinal responce into the code, something like
x=distance from wall on right;
y=distance from wall infrount;
z=100-((y+x)/10)
turn Z amount.
I made a wall follower once for university and the idea was it would always stay 3cm away from wall and follow it. i found the type of forumula above was fairly easy and produced good results,might make it more fluid.
if you have something like excel or matlab you can even find the best turning curves etc.
Anyways great project and i liked the videos alot.
Noz
Hum.. erh.. thanks..
Hum.. erh.. thanks.. hmm..
Well - to put it short; I think one thousand nerds out there have tried to make something like wall racers / Microcontroller on an old RC. But failed because they thought they could calculate the way out of things / predict etc like the robots they have experience from.
The thing is; These things (specially the cheap ones) steer like a numb fish, react like a drunk fish, and whip around like fish on speed in a heavy stream. It is NOT like a nice slow little robot. You cannot transmit formulas from "ordinary robots" that behave nice to these things, they will bang into the wall and act crazy if you try. It has to be formulas on how to steer a numb, drunk fish - on speed, in a heavy stream :)
But I will release both code and how-to's asap. I will release a very simple base-version that works (100% the one on the videos, and possibly more videos to come) - and then you can have all the fun in the world by trying different formulas - they take an afternoon to build, and are very cheap :)
Thanks again,
F.
A Fishy Tale
Walkthrough is comming -
Nice little critters
Realy like the fact that you have taken radiocontrolled cars and made them childproof, radiocontrolled cars that doesn't drive into thing that's definetly something I think a lot of parents would like and that way they also last longer
Have you thought about hidding the SRF05 sensors indside the car, can't quite come up with a way to hide ultrasonic sensors inside the cars, I can only think of a way to do it with infrared sensors.
The SRF's could just be
Cant you place the
I would assume that as long
Done something similar myself
I made some similar robots a while ago. I'll try and post a video of them. They are designed to race around a track, in other words they are doing wall avoiding so when you make a track with walls on the inside and outside they will go around and around. I made the robots myself, they are differential steering with a body made from laser cut polycarbonate. I used the sharp IR rangefinders for navigation and have four on each robot, two forward facing and two side facing. They also have four bumber switches, one pair in front and the other at the rear.
I made them because I occasionally run kids workshops about robots and I wanted an activity that was fun but didn't involve building or programming so I made some racing robots. The controller (A PIC 18 device) has eight potentiometers attached (as well as the sensors) and these can be used to control the behaviour. Basically I can give one to a group of kids and tell them to experiment with the settings and get the robot to race around the track as fast as possible. The controller also has a mode select switch so you can have several different operating modes and I have included (but not written code for yet) a pair of servo pulse outputs so you can use the controller on an RC car instead of a diff steer bot. I was thinking of also using these ports to hook up an RC helicopter gyro so I could control the rate of turn more easily (espescially useful on slippery floors)
My robots are FAST, not quite as fast as a good RC car but certainly to fast for most tracks (there is a speed setting so you can tune the motor power) They work best on carpet where they get the most grip. The algorithms I use for control are various and the mode switch allows you to try out different options. The most useful one is the PD controller where you make the robot symetrical with sensors at the front and sides in pairs. You then work out the difference ebtween each sensor in the pair, work out the rate of change of the difference, add the results together and subtract from one motor and add to the other. This gives you a damped wall avoiding behaviour which works nicely.
I haven't done what the wall racers do yet and given them an explicit mode for following a left or right hand wall, Mine just avoid walls, I think I got half way through writing some code for wall following and got distracted by other things.