Effectiveness of Casters

[briholton]

Originally Posted by KenWittlief

where are all the REAL engineers in this thread? ...

....you close the loop on steering by having the SW look at what the driver is commanding the bot to do

....and use a PID closed loop algorythm to MAKE the bot do what its told!

Hey Ken, can a REAL engineer point me in a direction to learn how to use the yaw rate sensor? I'm just a dumb physicist. I'm curious to its use and have seen it for a number of years now and always just piled it up in the dunno pile. Cuz, after all this talk, I have to give up on my casters..... darn it, had some really nice ones ordered too! hmm, perhaps for the goal.

[maxgebhardt]

We used castors our rookie year and had less qualifing points than a robot that didn't even move... they are awful. just imagine pushing a 130lb shopping cart at the supermarket. those castors are also tempermental and jam easily. if anyone on you team still likes castors, then take them to pathmark and you jump in a cart and tell them to push you for 2 minutes, and see how much excess energy it takes.

[KenWittlief]

I dont know if I can give enough info in a post to explain how to use them if your are not familiar with PID control loops.

The idea behind it is simple. If you are using single joystick control then the X axis is indicating how fast your driver wants to turn

if you just feed this into the single joystick equation that comes with the default code - then the stick postition is mapped directly to a PWM output, which ends up telling the motor how much force to apply (you can think of this as how much voltage is on the motor)

problem is - there is friction in your drivetrain that will resist movement - so putting a small pwm command on the motor will not be enough to make it move - once you have moved the stick enough to make it move, you have overcome the static friction and now you are dealing with kinetic friction, which is less, so once the motor moves a little, that pwm command is enough to make it move noticeably more

to make matters worse, the left and right drivetrains will not have exactly the same amount of static friction, so one will start sooner than the other. If you are telling the bot to go straight, it will goto one side a little.

The yaw rate sensor detects how fast the robot is turning - if you mount it with its bolt holes up and down.

the way a PID control loop works is, instead of taking the signal from the joystick and driving the motors pwm with that, you take the joystick X axis signal and you subtract the yaw rate sensor signal.

so if the driver is commanding a slight turn, and the yaw rate sensor says "im not seeing any turning going on here!" you end up with an error signal - the error is the difference between what the driver wants, and what the robot is doing.

the simplist form of closing the loop would be to use the error signal to drive the motors - replace the joystick X axis signal with the error signal and feed it into the one joystick equation. You have to adjust the error signal for 127 being zero, something like this:

yaw_error = (((2127 + p1_x - yaw) Min 2000 Max 2254)-2000)
yaw_command =(((2063 + yaw_error/2) Min 2000 Max 2254) -2000)

BTW - this is in basic from last year - wont take much to change it to C

In PID lingo this is a Proportional loop. The output to the motors is proportional to the difference between the commanded value and the actual (the feedback) value. how responsive the system will be depends on the gain. in the code example above, the error is divided by 2, so the gain is 0.5

thats not a bad place to start with a machine that is using two wheels and castors, esp if your weight is balanced towards the two drive wheels. if you try the code and the bot is sluggish, you can increase the gain ( and you have to adjust the 2063 number so that a yaw error of 127 (zero) stays zero

if you have a bot with 4 wheels, you might need more gain, like 2 maybe - in that case you would adjust the 2063 down to 1873.

if you crank the gain up too high, what will happen is, when the loop corrects itself it will over correct, and overshoot the turn rate - then the yaw rate sensor will say "you went too far" and try to correct it back, which will overshoot again, and the bot will twist back and forth. If the gain is just slightly too high, the bot will respond like its nervious and twitchy - if its way too high the bot will swing back and forth all by itself.

one thing you have to be carefull here - make sure your yaw rate sensor is mounted in the right orientation - for the code above we had our sensor mounted upside down, the holes were upward (simply because that was a convient place to put it on our frame). Also the sensor has to be solidly attached to the frame. If you tie wrap it on, and it can move around, it will go nuts on you.

so what are the I and D parts of PID? I is intergral - if you move the stick a little bit, the error signal might not be enough to get the motors to move - so to add an intergral part you take part of the error signal and you sum it, or accumulate it over many cycles - so if the driver is telling it to move, and its not moving, the error signal keeps increasing until it does.

The D is differential. sometimes you want the bot to respond quickly to changes in the joystick position, so you can either keep track of where the joystick was last time, and subtract where it is now, to tell when it has been moved, and apply a gain to that signal to sort of give the motors a 'kick' when the joystick has been moved - or you can use D on the error signal in the same way.

For our bot last year, all we needed was the P part of the PID loop - the equation I listed above. We ended up with the best response with the gain at 2, but I would try it at lower values and work your way up.

If you want to learn more about PID control loops, try searching the web - lots of companies make equipement that is used professionally, and I have seen some excellent websites that explain the concept better than I have here.

PID loops can also be used for thing like postioning an arm on the bot, or using pots to tell where your wheels are pointing if you use crab steering, or your front wheels turn left and right to steer.

hope this has been useful.

[KenWittlief]

I broused through a few web sites, if you search yahoo on

PID control feedback

you will get many links

this one gives a good overall explaintion without going into the math too much:

http://www.netrino.com/Publications/Glossary/PID.html

I found others that use fequency domain equations, if you can get past that, they are very informative - ie, skip over the math you dont understand and look at the graphs and you will get some insite.

[Spikey]

It all depends on what kind of competition you are competing in. In 2000 we built a casterbot that won 2 regionals, and in 2002 we built a caster bot that finished in the top 10 in both regionals, casters can be valuable, it all depends on the competition playing field.

[Gary Dillard]

Originally Posted by KenWittlief

the time it takes is one fourtieth of a second for the feed back loop to see that 'something' is trying to push the robot sideways, when its not suppose to be going sideways, and the SW corrects the power to the motor to force it to do what the driver wants.

This works to stablize castors, it also works if you hit a ball, if a bot hits you, you encounter a change in resistance (like driving on carpet or on the smooth surface)

and it worked last year going up the ramp at an angle.

The PID loop doesnt care whats trying to push the bot off course, it senses the slightest change in yaw rate, and applies an immediate correction.

We used it last year - the bot steered like it was a large servo.

At 10 fps, that's 120 ips, I've travelled 3 inches in 1/40th of a second. That's about 120 degrees of rotation on a 3 inch caster - it has already spun around and pushed me off position by then.

If you want to drive all over the field at part power because you're shutting down one motor or the other all the time to compensate for instability in your drive system, please do. We'd love to have you as an opponent.

Spikey, I'm impressed that you won 2 regionals with casters, but I'm betting it's not because of the casters right? You just focused your resources on other parts of the robot rather than on the drive system, so you had a robot that played all aspects of the game and didn't just drive.

The question of this thread is, what do you think of casters. My answer - not much.

[ShadowKnight]

liek alavinus posted, our team is team is using skids...the only thing that 4-wheel drive is good for is added traction...for manuevering stuff liek the steps, it'd be great...6 could work but I'm not entriely sure that many are needed. I'm not entirely sure that you would WANT to get onto the stairs though because if I read your team number correctly, you're a a rookie team...having been on a rookie team myself and knowing a number of people who were rookies at one time...I would focus on a driving box...maybe something that herds balls...and use 1 caster not 2 if your team decides to use casters, though I like many others do not recomend it.

But like I was saying, our team is using skids...so the front of our robot will just slide along the surface of the carpet. you guys might want to look into that...

[KenWittlief]

if you dont like castors then the next best thing is to have steerable wheels, like a car or wagon.

4 wheel drive with tank/skid steer is hidious - the amount of power and energy you waste dragging a drivewheel sideway is tremendous - in this years game, as other have pointed out, if you dont want to use castors, use a couple blocks of nylon instead, and have most of your weight balanced over your drive wheels.

[psychoCHIPMUNKK]

I have to agree with minic@HYPER69, castors are not a good idea. Imagine having a 130 lb robot hit you when only 2 wheels will resist turning. Last year there were a few robots that had castors that tried to make it up the ramp, and the only ones there were able to do it were the ones that were smart enough to drive so their powered wheels went first up the ramp. A 4 wheel design works just fine, and a 6 wheel can get a bit tricky if you don't know what you're doing. As it's been said before, having castors as part of your drive train is the easiest way to identify a rookie team.

This team tried to make it up the steps using a 2 drive/2 castor robot, it didn't work
Here

[Arefin Bari]

sorry if this was mentioned in this thread before...

one good reason that you dont want to put casters on because during autonomous period you will lose your track. i have real bad experience with casters and i would like to suggest to not to use them as everybody did in this thread...

[Al Skierkiewicz]

Although we have used casters in the past, we have decided that for most applications, a caster design will be considered last. If you are moving in one direction and want to reverse, the casters have a mind of their own as to which way to turn and rarely will they allow you to back up in a straight line. For last year's competition we had four wheel crab but did find it neccesary to turn the robot at times. For this we had a pnuematic cylinder drop a "foot" to lift two wheels off the ground. This allowed a change in attitude and moved the center of turn to one side of the robot. It is tricky for drivers to learn but with practice it was very effective.

[Raul]

Originally Posted by Al Skierkiewicz

Although we have used casters in the past, we have decided that for most applications, a caster design will be considered last. If you are moving in one direction and want to reverse, the casters have a mind of their own as to which way to turn and rarely will they allow you to back up in a straight line. For last year's competition we had four wheel crab but did find it neccesary to turn the robot at times. For this we had a pnuematic cylinder drop a "foot" to lift two wheels off the ground. This allowed a change in attitude and moved the center of turn to one side of the robot. It is tricky for drivers to learn but with practice it was very effective.

Let me add to that. We had casters in 1999 when we had 2-wheel crab. It was terribly hard to drive straight when the casters were trailing. We will NEVER use casters again. Omni-wheels are much more stable!

[Ryan M.]

If you're thinking about getting onto the platform at all your casters will have to be incredibly strong. Even the small steps will have a tendency to snag on them and the casters will very likely be ripped off.

[Ed Sparks]

Originally Posted by Solace

although It's possible one could make a casterbot that worked reasonably well, I can tell you from experience that looking for a 2 wheel 2 caster drive train is in general the easiest way to pick out the inexperienced teams from the crowd. The casterbots that manage to retain even a mediocre level of competitiveness are few and far between.

4 wheel drive works fine. 6 wheel drive works better still.
In short, avoid casters AT ALL COSTS. They will be your downfall.

I disagree ........

We are a very experienced team (34), we will be utilizing some very cool casters, we will be on top of the platform, and we WILL be competitive.

You just have to think outside that norm a little ........

[Gary Dillard]

Originally Posted by Al Skierkiewicz

It is tricky for drivers to learn but with practice it was very effective.

Oh Al, you master of the understatement. Your robot ROCKED last year!!! As they do every year.