Sensor To b used to Sense a pit : URGENT PLS

[evilknight007]

Hey Ppl .

 I'm Bout to build a autonomous robot for a competition. The Problem Statement for the contest can b seen in the link below.

http://www.shaastra.org/n/events/Mission+Mars

Is it better to use a tread wheel or a Four wheel drive ..and also pls tell how to avoid the pit and find the bridge.

I'll need all the help i can . Pls help me.

Thank U all .

[Tsukubadaisei]

That i a very cool contest.
I am participating in one very similar right now. Actually my robot is already built if you are interested I can post photos and info.
I would go for 4 wheeled with dual steering with pneumatic tires like I do most of the times, but treads will do just as fine or even better because of the rocks.
Since your robot mus be small (25X25X60cm^3) I think you must use a camera. Another option is making  "nose" with  few photoswitches but that is out of question because of the rules. If the arena is exacty like in the 3D model then you wont have that much of problem for detecting the pit.

So, if I had to build a robot for this contest I would do for 2 stepper motors, treads a camera and a small computer or umpc.

Please ask any question you want, I am very interested. As I said before, I am facing a similar contet right now.

[evilknight007]

Tat's So nice of u .Thanks 4 ur support.

But i'm a complete novice and could do with some more explanation.

I'd be interested to see ur robot's photos.

[Admin]

These should get you started:
http://www.societyofrobots.com/toparticles.shtml

Sounds like you want a chassis similar to this:
http://www.societyofrobots.com/robot_asme.shtml

[KambeiX]

Since you are a beginner, try to keep it simple


As I told you before, you can use some line following sensors pluse some mechanical sensors.

For pit sensing you can try something like this:

http://www.acroname.com/robotics/parts/R185-SINGLE-LINE-IR.html

or build your own (using infrared emiter-reciever diodes or modules like cny70)

and for rocks you can detect them using a bumper switch like this

http://www.robotstore.com/store/product.asp?pid=1021&catid=1601

or, of course, you can build your own for less.

[evilknight007]

Thank u ppl .. i hvn't got this much support from anyother forum. Amazing !!

I have a doubt here . The arena surface is rough  . Will Ir sensors be able to reflect any light to create a output.

Pls keep posting . I'll buzz u if i hv anything more to ask.

[KambeiX]

they will because the threshold is very wide (25cm pit vs a few cm sand)

[Tsukubadaisei]

Here are the pictures. The first is a CAD and and the others are the real version. It is going to participate in 2 contests similar to yours (transport an apple through a rough terrain with hundreds of obstacles without touching it). The robot is 70% done. In August I will post the complete version of it with all the sensors and everything connected. I was planning writing a tutorial about it for the SoR2007 but unfortunately I have another project in mind just now and it is almost not related with robotics at all, so I am busy.







The CAD and the real thing are a bit different because I made a few upgrades in the plan while building.

I disagree with KambeiX about using sensors because:

1) As I said before the robot must be small and when you use a linetracer you need a "nose".

2) if you are a beginner I dont recommend sensing the rocks because it will make the source code complex and, if use treads, you may just pass over them.

3) Don't buy those sensors, they are very expensive. you can build them following on line guides for just a few cents.(if you want to use sensor)

I recommend using a computer with a webcam and a freeware program called roborealm for the vision. Then you connect your robot to some controller that can do PWM(pulse width modulation, pulse used for controlling motors, for example) and then connect the controller to the stepper motors(High torque servos will work as well, I am using them). S.Motors are good because of the angle feedback, you have absolute controll and they usually have high torque.

And read everyting ADMIN wrote in those links. And don't forget that Google is your friend.

[evilknight007]

 k man . . I'll c to it .

[novice88]

To Tsuku-

I'm doing this project with evilknight... using a computer is outta the question as i) its too bulky if placed along wit the bot and ii) we are not suppose to control the bot in anyway after it has entered the arena... we are thinking of using a microswitch with spoons... do u think this would be feasible? or do u have any other suggestions? please advice..

[KambeiX]

microswitches as bumper sensors are cheap and self made sensors are cheap too, you just need to know that the signal you are emitting (light or IR) isn't getting back (pit).

I thought of tank threads too but rocks will be >7cm so what if they are 20cm high? max wheel diameter is 10cm so that limits you on what you can pass.

Of course there can also be other approaches like using a range finder instead of line following sensors. In fact, this would be better but implies the use of a range finders which require more programming and are more expensive.

Anyways, you could do this, mounting a range finder on top of a servo and when you run close to the pit, make the robot turn and aim the range finder to the pit until it finds a bridge, then turn, or better yet, put a range finder on each side of the robot and one in front and do the same, of course you will spend a lot so that's why I proposed the line follower idea, that acts as a proximity sensor, the same as the range finder only a lot cheaper and more simple.

[Tsukubadaisei]

To Tsuku-

I'm doing this project with evilknight... using a computer is outta the question as i) its too bulky if placed along wit the bot and ii) we are not suppose to control the bot in anyway after it has entered the arena... we are thinking of using a microswitch with spoons... do u think this would be feasible? or do u have any other suggestions? please advice..

I still think the computer would be the best choice:
i)There are really small computers (15cmX15cm)
ii)If you use a computer you dont need to control the robot. The computer IS the brain of the robot.
I mentioned a computer because is easier to program a camera-computer-robot than a camera-controller-robot.
Of course, controller-robot is much lighter, and that would be an advantage.

For controllers my suggestions are Freescale 8bit controllers(for $70 you get a powerfull controller, a development environment for C called Codewarrior and the compiler and it is available in anywhere in the world) or H8 from Renesas/Hitachi if you manage to import from Japan. Another good enough option is PIC, but I have never used. Freescale controllers are not very popular between hobbyst, so you will not find a lot of stuff on the net. I only use them because they are one of my sponsors. H8 is very popular where I live, when you buy it it comes as a complete set like freescale but most of its comunity only speaks Japanese which I think is out of question for you guys. PIC on the other hand is not so user friendly for beginners but it has a huge comunity in US and EU. It is very easy to find sample programs for it. Even this site has some.

For sensing then I would go for 3 sharp PSD sensors gp2d12 (the "range finder" as KambeiX said, but it is not the correct way to call it, I cant understand why everybody in this board call PSD range finders, the function might be the same but they are pretty different devices, range finders are very expensive). I would put one in the front and two on one side of the robot and all of them would face towards the ground. They would keep a certain range of values until one of them get over a pit. In this moment the reading in the front sensor will change widely. That will tell your robot that there is a pit. Then the robot will turn left or right, it depends where are your side sensors, and with the side sensors it will follow the pit like if it was following a line.

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
xxxxxoxxxxxxxxxxxxxxxxxxxxxxxx <-----pit
-------+------------------------------------
       |       o <----sensor
     [][][][][][][]
  o__|          |
       |          | <----robot
     [][][][][][][]<----treads

Try to put your sensor like in the above ASCII diagram, in such a way one sensor will be facing the pit and the other the ground. Program your robot so that it will always keep one sensor over the pit ond the other over the ground. If you succeed in doing this you will not only manage to not to fall in the pit but also will find the bridge and pass it.

The next problem will be stooping the pit following routine after passing through the bridge. For that you could use the stepper mottor. That is very simple: you have to make your robot turn about 90 degrees in order to enter the bridge. If you succeed writing the above program your robot WILL pass the bridge. So the only thing you have to do is write another subroutine that will detect that your robot rotated about 90 degrees by counting the number of PWM pulses the controller sent to the motors/motor controller(you need one: google for h-bridge). If the function returns true it will stop the pit follower function and the robot will just go forward to the exit.

Those are the basics of my solution. Of course there a other solutions. There is nothing more I can say. Everything I mentioned here you can easy find on the net how to build or control. I think now it is up to you. I don't think this contest is THAT difficult at all so a simple robot may be easily able to succeed (even simpler that I suggested) if you are good enough with C or maybe just luck will be enough(believe me luck is a huge factor when building your very first robot).

[evilknight007]

Thank u mate. 
 
Tat info would be fine . But there's one more thing . The bridge need not be straight as given in the problem statement .

so the line following routine should still be running even after the detecting the bridge. . ii'll try to implement it. Thank you both..this has been great help.

[KambeiX]

To Tsuku-

I'm doing this project with evilknight... using a computer is outta the question as i) its too bulky if placed along wit the bot and ii) we are not suppose to control the bot in anyway after it has entered the arena... we are thinking of using a microswitch with spoons... do u think this would be feasible? or do u have any other suggestions? please advice..

I still think the computer would be the best choice:
i)There are really small computers (15cmX15cm)
ii)If you use a computer you dont need to control the robot. The computer IS the brain of the robot.
I mentioned a computer because is easier to program a camera-computer-robot than a camera-controller-robot.
Of course, controller-robot is much lighter, and that would be an advantage.

For controllers my suggestions are Freescale 8bit controllers(for $70 you get a powerfull controller, a development environment for C called Codewarrior and the compiler and it is available in anywhere in the world) or H8 from Renesas/Hitachi if you manage to import from Japan. Another good enough option is PIC, but I have never used. Freescale controllers are not very popular between hobbyst, so you will not find a lot of stuff on the net. I only use them because they are one of my sponsors. H8 is very popular where I live, when you buy it it comes as a complete set like freescale but most of its comunity only speaks Japanese which I think is out of question for you guys. PIC on the other hand is not so user friendly for beginners but it has a huge comunity in US and EU. It is very easy to find sample programs for it. Even this site has some.

For sensing then I would go for 3 sharp PSD sensors gp2d12 (the "range finder" as KambeiX said, but it is not the correct way to call it, I cant understand why everybody in this board call PSD range finders, the function might be the same but they are pretty different devices, range finders are very expensive). I would put one in the front and two on one side of the robot and all of them would face towards the ground. They would keep a certain range of values until one of them get over a pit. In this moment the reading in the front sensor will change widely. That will tell your robot that there is a pit. Then the robot will turn left or right, it depends where are your side sensors, and with the side sensors it will follow the pit like if it was following a line.

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
xxxxxoxxxxxxxxxxxxxxxxxxxxxxxx <-----pit
-------+------------------------------------
       |       o <----sensor
     [][][][][][][]
  o__|          |
       |          | <----robot
     [][][][][][][]<----treads

Try to put your sensor like in the above ASCII diagram, in such a way one sensor will be facing the pit and the other the ground. Program your robot so that it will always keep one sensor over the pit ond the other over the ground. If you succeed in doing this you will not only manage to not to fall in the pit but also will find the bridge and pass it.

The next problem will be stooping the pit following routine after passing through the bridge. For that you could use the stepper mottor. That is very simple: you have to make your robot turn about 90 degrees in order to enter the bridge. If you succeed writing the above program your robot WILL pass the bridge. So the only thing you have to do is write another subroutine that will detect that your robot rotated about 90 degrees by counting the number of PWM pulses the controller sent to the motors/motor controller(you need one: google for h-bridge). If the function returns true it will stop the pit follower function and the robot will just go forward to the exit.

Those are the basics of my solution. Of course there a other solutions. There is nothing more I can say. Everything I mentioned here you can easy find on the net how to build or control. I think now it is up to you. I don't think this contest is THAT difficult at all so a simple robot may be easily able to succeed (even simpler that I suggested) if you are good enough with C or maybe just luck will be enough(believe me luck is a huge factor when building your very first robot).

Your solution is very good indeed. I just suggested a cheaper alternative.