Newbie, special needs for car motor/control

[jpwinter]

Greetings all!
Okay, yes, I'm a newbie. Please be kind.
That being said, I have a special need for a motor, and I'd like to go brushless... So, here is what I'm doing:
I have a HPI Pro4 (hara) edition kit I just built, and I need a motor and controller for it.
This is not for your average racing! I am building a robotic car for a robot race (interested in details? check out http://www.cs.cmu.edu/~mobot/ ), and as such, my needs are a bit different than your average go-as-fast-as-you-can car. I have no radio controller. I need to go fairly fast, but don't have straight aways for long accelerations, and have lots of curves (4' diameter circles). I'll probably need ideas for a good steering servo, as well... I currently have a standard hitec, but since I'm using bigger/wider tires, that may not suffice.
I've been looking at the plethora of motor/controller/battery designs, and rather than just pick something out of the hat, I'm asking you experienced people for a few suggestions!
I've changed a couple of things on the car so far: I'm increasing the wheel diameter up to 2.2" to get some more ground clearance, as I need to mount some sensors/electronics under the car to help with the guidence. Hmmm... what else might be of interest and help you make suggestions... I have no problem making parts if I have to, I have a cnc mill, so can make adapters if necessary. I figure I'll be using a very small pinion gear with the standard (so far) gear on the drive shaft; I expect my max speed to be slower than most everyone else would expect.
The course is a sidewalk in front of the CS building; total course is about 300 feet, with a vertical drop of almost 8 feet (in two ramps, one 4' tall and about 8' long, i.e., about 30 dergree ramp) and the other about 3.5 feet tall, same ratio... Being a sidewalk, it has gaps and bumps that we have to run over, hence, the larger tires to get me over the uplifts.
If that's not enough info, I'll be happy to provide more.
If this is not something that is of interest, please tell me and I'll not bother you anymore. :**

Thanks in advance, I really would appreciate constructive suggestions for controls/motors ways to improve handling so I can speed up!

Ciao!
Jason

[tcolesen]

This is an interesting project! A sensor driven non-RC car would be awesome to see in action.

There would be many choices for a brushless system for a project like this. To start out with, you could get a Mtroniks Pro controller. This can handle just about any motor you could throw into a touring car. For your purposes, it will be more than enough. For a motor, you could get a Feigao 380C 13t. The whole combo will cost ~$225.

Then there are the batteries. You can start out with lightweight LiPo batteries, or you can go with NiMh. If you wish to use LiPos, I would suggest you use a 2s (2 cell series, 7.4v) battery with the above recommended motor. On regular gearing, this would give you about 20mph. You could gear lower, but the problem with gearing too low is that the motor will be running so inefficiently (not enough load) that it will get hot. As for capacity, anything would work. You could get a battery with 8000mah for LONG runtimes, about an hour (possibly more).
If you want to go the cheaper (but heavier) route, you can use 6 NiMh cells (7.2v). GP3700 or IB3800 would be the top picks for cells.

I am really interested in what you are doing with the sensors on a vehicle like this!

[tcolesen]

Another possible option would be to use a sensored motor system. This may prove to be more reliable for such a vehicle. If you were to choose this route, you could use a Novak SS4300 system on 4 NiMh cells. There is no doubt that it would be pretty slow.

[jpwinter]

This is an interesting project! A sensor driven non-RC car would be awesome to see in action.

It's getting fun! I'm designing the sensor boards for under the car, it'll have a gyro for sensing the turning force, and possibly a few sensors to determine when I hit bumps (like the joints in the sidewalk) to help detect position anomolies.

There would be many choices for a brushless system for a project like this. To start out with, you could get a Mtroniks Pro controller. This can handle just about any motor you could throw into a touring car. For your purposes, it will be more than enough. For a motor, you could get a Feigao 380C 13t. The whole combo will cost ~$225.
How about the Novak Super Sport with the SS5800 motor? reading over the many different combinations, that on seems to be reasonably good, and I found it for 229 as well...

Then there are the batteries. You can start out with lightweight LiPo batteries, or you can go with NiMh.

I think I'll go with the LiPos, as the lower weight should help out. Besides, I'll be adding weight in the form of cpu's (several, networked together) and another battery to power them. So I'll reduce weight where I can.
If you wish to use LiPos, I would suggest you use a 2s (2 cell series, 7.4v) battery with the above recommended motor. On regular gearing, this would give you about 20mph.

20 is about what I think is max; I don't think it'll stay on the ground any faster (if even at that speed!), so that sounds great! I'm still not upto speed on things like 12c, 15c, 2s, etc.. is there anywhere I can get info on what the battery combinations do, and why they are good for certain things?

You could gear lower, but the problem with gearing too low is that the motor will be running so inefficiently (not enough load) that it will get hot.
That doesn't sound right... that would say that you should never free run a motor? And free running a motor doesn't kill them, at least, not other motors. Are these different for some reason? Or is there something I'm not seeing?

I am really interested in what you are doing with the sensors on a vehicle like this!

Me too! Under the car, are three banks of line sensors; front, middle and back. That way, I can use the curvature of the line to help me know when I'm on it, vs. when the backend has slid sideways and I'm not paralell to the line anymore. The g-sensors are mounted vertically, and will be able to tell when it hit's a uniform bump, such as the joint in a square. Finally, I can use the gyro to read the turning force, and use it to help figure out how fast I can take corners for a particular condition. I think. :eek: Since the race is held no matter what the weather, it has to be able to adjust to the "track" conditions. It's fun, and the more I work on the design, the better it gets! I've got multiiple cpu's handling different facets of the sensors, so the main cpu doesn't have to worry about pre-conditioning information, just using it. Makes the main code much simpler.

Thanks for the help and advice!

Ciao!
Jason

[tcolesen]

It's getting fun! I'm designing the sensor boards for under the car, it'll have a gyro for sensing the turning force, and possibly a few sensors to determine when I hit bumps (like the joints in the sidewalk) to help detect position anomolies.

I would have never thought of using a gyro!

How about the Novak Super Sport with the SS5800 motor? reading over the many different combinations, that on seems to be reasonably good, and I found it for 229 as well...

The Novak Super Sport with 5800 may be too fast. If you use Lipos, you will have to go with a 2 cell pack, which is 7.4v (nominal). The motor might get too hot from inefficiency on a higher voltage like this. That is why I suggested the 4300. BTW, you can get the Novak systems for a lot cheaper than that:
4300 system (http://www.shopatron.com/product/product_id=NOV3006/135.0.151.0.0.0.0)
Remanufactured 5800 system (http://www.shopatron.com/product/product_id=NOVR3005/135.0.7565.7521.0.0.0)


I think I'll go with the LiPos, as the lower weight should help out. Besides, I'll be adding weight in the form of cpu's (several, networked together) and another battery to power them. So I'll reduce weight where I can.

For what you will need, Lipos can also be smaller.

20 is about what I think is max; I don't think it'll stay on the ground any faster (if even at that speed!), so that sounds great! I'm still not upto speed on things like 12c, 15c, 2s, etc.. is there anywhere I can get info on what the battery combinations do, and why they are good for certain things?

The different "C" rates will tell the maximum rate at which a Lipo can be discharged. For example, a 2000mah 15C Lipo can safely be discharged at 30amps, or a 4000mah 10C at 40amps. These are continuous ratings, however, and usually there will be a burst rating. The burst rating would be a higher sustained amp draw for about 10sec. that the pack would be capable of handling. If you want to really get into Lipos, you can visit the "Battery and Charger Forum" at www.rcgroups.com

That doesn't sound right... that would say that you should never free run a motor? And free running a motor doesn't kill them, at least, not other motors. Are these different for some reason? Or is there something I'm not seeing?

A brushless motor should never be allowed to rev high without a load. Usually the kv (RPM/volt rating) that a motor is given is under some sort of a load, so it would be easy for it to spin past the RPM rating of the bearings, thus damaging the bearings. Also, at high RPMs it is possible for the magnet to be released from the shaft, which could possibly destroy the coils in the motor. It is safe to run a motor on low throttle without a load, though.

Me too! Under the car, are three banks of line sensors; front, middle and back. That way, I can use the curvature of the line to help me know when I'm on it, vs. when the backend has slid sideways and I'm not paralell to the line anymore. The g-sensors are mounted vertically, and will be able to tell when it hit's a uniform bump, such as the joint in a square. Finally, I can use the gyro to read the turning force, and use it to help figure out how fast I can take corners for a particular condition. I think. :eek: Since the race is held no matter what the weather, it has to be able to adjust to the "track" conditions. It's fun, and the more I work on the design, the better it gets! I've got multiiple cpu's handling different facets of the sensors, so the main cpu doesn't have to worry about pre-conditioning information, just using it. Makes the main code much simpler.

Thanks for the help and advice!

Ciao!
Jason

Your array of sensors shouldn't be able to miss the line! I watched the 2005 video, and some of those "vehicles" would track the line, but then miss it. Your way sounds foolproof.

[SpEEdyBL]

I don't see what's wrong with the 380c 13t. Motors don't heat up unloaded unless they are pulling a lot of amps. Low turn motors such as a 540 6s or 380 6s have unloaded heat problems because they over rev and will pull a lot of amps. At 7.4 volts, the 13t is no where near over reving and doesn't draw that much current at no load.

[kufman]

Sounds like a sensored system is what you need. For this, I am not sure why you would go with brushless at all, but sensored is that way to go for precise control.

[jpwinter]

Sounds like a sensored system is what you need. For this, I am not sure why you would go with brushless at all, but sensored is that way to go for precise control.
Sensored sounds about right, and as for why brushless: Well, I've been reading the boards, and all of the stuff for brushes seems a bit daunting, when all I want it to do is drive.. :) Seemed to me that brushless would cause less problems over the long run, and let me concentrate on what I really wanted to do; make it autonmous and fast.

I suppose in reality I don't *need* brushless, but I'll go ahead and get it. Who knows what I'll be doing next time with this? I may be switching to the faster motor, and use the gyro to run a counterbalance to keep me on the ground through corners much too fast. Ya gotta dream to do things!

Thanks!

Jason

[GordonFreeman]

I would just go with a no-nonsense Novak BL system. Don't waste your energy trying to figure out the exact best power sytem for this app. Use your energy and time to build the best controller (CPU/software) for the bot as that will probably be the limiting factor, not the car speed. Looking at the few pictures on the site I would be impressed to see a robot car follow that line nearing 20 MPH! Without some "look ahead" or previous knowledge of the track, it's going to overshoot some of those corners if it is going 20 MPH when it needs to turn in. Keep us posted, I wanna see a video!

[jpwinter]

I would just go with a no-nonsense Novak BL system. Don't waste your energy trying to figure out the exact best power sytem for this app. Use your energy and time to build the best controller (CPU/software) for the bot as that will probably be the limiting factor, not the car speed. Looking at the few pictures on the site I would be impressed to see a robot car follow that line nearing 20 MPH! Without some "look ahead" or previous knowledge of the track, it's going to overshoot some of those corners if it is going 20 MPH when it needs to turn in. Keep us posted, I wanna see a video!

20mph is a dream; I'm aiming for 10 to start. :) The current course time to beat is 1 minute 6 seconds. I think we calculated that 10mph would almost half that time.
As for look ahead; we do have that; infact, we get the course layout 24 hours in advance, so we can put the gates into whatever mapping scheme we are using. The real idea is to go through the gates in order, not just to follow the line.
I'm following the line this year just to get it going; then, next year, I'll add vision and gate recognition, and aim to drive between gates and to heck with the line... :eek: Well, okay, I'll still probably use the line when nearing a gate to make sure I go through in the center. Knocking over a gate disqualifies you.. :cool:

Thanks! I will post when the new video is out, or when I make one of my own to show what I've been doing!

Ciao!

Jason

[Jev]

Awsome project, I can see some great scope for artificial intelligence :). IMHO the only way to go real fast would be with forward facing visual guidance. The car would need to be able to predict as far in advance as possible the next section layout so it could position the car better on coner entry. It sound like you are using a closed loop program this time, at least you can spend time sorting the chassis rather than all the time on code.

If you wanted to go faster without sticking rigidly to the white lines then you need a system to identify when the car is aproching a gate. Would you be able to use radar? :)

Good luck

[jpwinter]

... IMHO the only way to go real fast would be with forward facing visual guidance. The car would need to be able to predict as far in advance as possible the next section layout so it could position the car better on coner entry. It sound like you are using a closed loop program this time, at least you can spend time sorting the chassis rather than all the time on code.

True, right now, it's basically a closed loop. Sensors, dead reckoning encoders, and gate sensors to help reset position. All to help tell me where I am on the map.


If you wanted to go faster without sticking rigidly to the white lines then you need a system to identify when the car is aproching a gate. Would you be able to use radar? :)

Good luck
Yes, radar would be cool! Probably overkill, for the range given though. I've got my eye on a camera for the next version (or sooner, if I can afford it). We're doing video processing in one of my classes. It would take multiple filters to pull the line out of the noise, but hey.. that's why you have multiple processors! Run them as sequential inline filters, and you can process full speed even with cheap/low power processors. <grin>
A lot of people have tried sonar, and some are talking about using inductance (the gates are soft steel) to tell when they pass a gate. The best time so far was done using vision, following the line. Nobody has yet finished by not following the line. So, that's for next year. Line first, then use vision to just find the gates and dodge through them!

That's going to be tough, though. Especially with the crowds around the edges, and the unreilable lighting, etc, etc. Fun, though! can't wait to get it up and running! Next up: designing and getting the sensor/cpu boards fab'd.

Thanks Guys!
Ciao!

Jason