Shop OBEX P1 Docs P2 Docs Learn Events
Idea: Robot with powered caster wheels, two independently powered wheels per c — Parallax Forums

Idea: Robot with powered caster wheels, two independently powered wheels per c

Dennis FerronDennis Ferron Posts: 480
edited 2009-11-15 22:58 in Robotics
I know some researchers at the University of Oklahoma who work on a six-wheeled robot chassis similar in design to the Mars rovers. (Six powered wheels, each wheel being able to turn left/right 360 degrees so the robot can crab sideways, spin in place, or steer in any direction.) It takes a lot of torque to turn those wheels left/right though. The first time they tried the robot they burned out all the servos pretty much instantly because they couldn't handle the torque needed to steer the wheels. Even if they manage to find servos big enough to do it, I'm not sure the 1/4 inch aluminum input shafts won't shear under the stress!

That got me thinking: rather than try to engineer something that can take the stress of steering those wheels, why not make it so that the drive motors can also steer?

What I'm talking about is use double caster wheels (imagine the way the front landing gear wheel for a jumbo jet looks) and make it so that each side is independently powered. By turning the wheels on each side of the caster in opposite directions, you could make the assembly turn on its own. And you'd want to offset it slightly (caster wheel, like the front wheels on a shopping cart) so that when the wheels are pulling the robot they'll tend to self-align with all the casters facing forward, and if a motor goes dead it will tend to self-align with that caster being dragged.

I notice Pololu has double gearboxes on sale for ~$9, so it wouldn't be very expensive to build. I was thinking of 3 casters to start out (would make a tripod robot) but then again if I want to be able to demonstrate it pulling a dead wheel I might want more total casters than 3...

With the gearboxes so cheap I would say the biggest expense would be the motor controller and HERE's where the self-aligning caster concept can shine: once all the casters are pointed in roughly the same direction, you should be able to drive all the motors together on the same controller. And you only need to align them one at a time. So if you use MOSFETs to selectively break the circuit between all but one of the casters' motors at a time, you could use a single dual motor controller to align all the casters, and then once you're happy with the positions use a single high current driver to drive all the motors together in the forward direction.

Comments

  • P!-RoP!-Ro Posts: 1,189
    edited 2009-11-06 22:05
    University of Oklahoma is working on larger robots now? I've only seen them work with smaller robots. Working on my own robot with similar steering requirements I've come across problems of too little power myself. I still have to test it but I'm now using a system similar to a linear actuator that I built myself which will allow me to now adjust the torque I apply. It would probably steer better if I used differential steering as you mention, but it would have made the mechanics too difficult as I am using a large engine to power it all. It will be interesting to see how this works out, robot mechanics has always been my primary interest. It's nice to see them now working on this sort of thing.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    PG

    Post Edited (Pi Guy) : 11/6/2009 10:12:55 PM GMT
  • Dennis FerronDennis Ferron Posts: 480
    edited 2009-11-06 22:14
    Oh well this robot was actually started several years ago; when I saw it about 3 years ago the project was temporarily stalled while they tried to decide what to do about it. In the time since they may have made lots of progress, or the same - I haven't been by the lab recently to look.

    But coincidently OU *is* actually building some large robots. There was a video posted to hackaday.com several weeks ago of a large gasoline-engine robot someone's working on. I'm not personally acquainted with the members of the group doing that robot, but I recognized having seen some of the parts in the lab, before they became part of a robot.
  • ercoerco Posts: 20,256
    edited 2009-11-06 23:28
    Novel idea, multiple diff-drive casters. You'll need a rotary encoder on each bogie to know what direction it's heading in. It might work well enough on a smooth & flat floor. Just like regular differential drive robots, the two motors will eventually curve off from a straight heading and will need constant encoder monitoring and course correction. Likewise, any bumps encountered will probably cause the individual casters to deflect and require course correction. You'll still need 2 motors per bogie/caster wheel and one rotary encoder for wheel direction, possibly one or more encoders per caster to measure distance or monitor wheel slippage. So it might end up being more complex than the multiple ganged steering bogie design you first mentioned. That design can use just one motor and gears or chains to sync & steer all wheels, and as little as one other motor or one motor per wheel for driving.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    ·"If you build it, they will come."
  • TimmooreTimmoore Posts: 1,031
    edited 2009-11-06 23:53
    If you chain the bogies together so they all point in the same direction, then you only need 1 encoder for direction. Then 1 of the bogies needs 2 motors to turn the bogie the other bogies will turn via the chain. Or have 2 of the bogies with 1 motor on different sides, with the otherside free-spining wheel. Those 2 motors work together for turning the bogie and drive motors i.e. different direction on the motors turn the bogie, the same direction drives the bot. The 3rd bogie doesn't need motors at all. You will have to angle hte bogie slightly to allow for friction/CoG from the 3rd bogie depending on the angle from bot direction to where the bogies are.
  • Dennis FerronDennis Ferron Posts: 480
    edited 2009-11-07 02:05
    Problem with chaining them together is to crab you'd want them to steer in the same direction but to spin in place you'd want them to steer in opposite directions.
  • P!-RoP!-Ro Posts: 1,189
    edited 2009-11-07 04:38
    Dennis Ferron said...
    Oh well this robot was actually started several years ago; when I saw it about 3 years ago the project was temporarily stalled while they tried to decide what to do about it. In the time since they may have made lots of progress, or the same - I haven't been by the lab recently to look.

    But coincidently OU *is* actually building some large robots. There was a video posted to hackaday.com several weeks ago of a large gasoline-engine robot someone's working on. I'm not personally acquainted with the members of the group doing that robot, but I recognized having seen some of the parts in the lab, before they became part of a robot.

    You ought to show me the link to that gas-powered one, for some reason I'm strangely interested in that. I've been to Norman many times and have seen robotics articles from that school but never saw anything larger than a foot long, so I guess that's why I came to that conclusion. I guess if they have anything good enough I might even go there after high school, but I'm not worried about the future right now.

    Back to the subject, there is one other way past differential drive, or an addition if you wish. What if the wheels were offset from their upper mounting and attached by a pivoting joint. This joint would have a brake on it, but can be released at times, allowing that wheel to rotate as it turns. Once at the desired point, it could once again lock and keep it in place. A bit tedious, but it could possibly become an advantage on the robot for moving around obstacles. Just an idea of many for the pool, but I'm sure every one of them is good in their own way. It just depends on what tasks the robot is required to do.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    PG
  • Dennis FerronDennis Ferron Posts: 480
    edited 2009-11-08 00:51
    Pi Guy said...
    Dennis Ferron said...
    Oh well this robot was actually started several years ago; when I saw it about 3 years ago the project was temporarily stalled while they tried to decide what to do about it. In the time since they may have made lots of progress, or the same - I haven't been by the lab recently to look.

    But coincidently OU *is* actually building some large robots. There was a video posted to hackaday.com several weeks ago of a large gasoline-engine robot someone's working on. I'm not personally acquainted with the members of the group doing that robot, but I recognized having seen some of the parts in the lab, before they became part of a robot.

    You ought to show me the link to that gas-powered one, for some reason I'm strangely interested in that. I've been to Norman many times and have seen robotics articles from that school but never saw anything larger than a foot long, so I guess that's why I came to that conclusion. I guess if they have anything good enough I might even go there after high school, but I'm not worried about the future right now.

    Back to the subject, there is one other way past differential drive, or an addition if you wish. What if the wheels were offset from their upper mounting and attached by a pivoting joint. This joint would have a brake on it, but can be released at times, allowing that wheel to rotate as it turns. Once at the desired point, it could once again lock and keep it in place. A bit tedious, but it could possibly become an advantage on the robot for moving around obstacles. Just an idea of many for the pool, but I'm sure every one of them is good in their own way. It just depends on what tasks the robot is required to do.

    Here's the link to the gas powered robot at OU:

    hackaday.com/2009/10/06/atv-brings-skynet-closer/

    Whereabouts in Oklahoma are you from, Pi Guy? I live in Moore.

    PG, that's an excellent idea on the locking pivot. It would work for some of the powerwheels robots I build. I read a whitepaper somewhere about a kind of adaptation of differential/skid steering where a large four-wheeled robot uses servos to steer the wheels slightly inward while executing turns to get a more favorable angle. (Otherwise with four wheels they skid while you're doing differential steering.) I thought, why use powered steering to do that? Why not, as you say, offset the wheels somewhat and allow them to pivot freely within some range. But of course the problem with that would be that the wheels would wobble while the robot tries to drive straight. But if you could lock the pivot when you want then that's a moot point. I don't think it would even be something you have to stop the robot and wait on: if we're just talking about pivoting the wheels within a very limited angle, like 30 degrees, you could get the wheels around to the most favorable angle for the turn you're going to make almost simultaneously to executing the turn, by giving some wheels more or less power, and then lock in place.

    For the pivot locking mechanism, I'd wrap steel aircraft cable 8 or 10 turns around the pivot axle and use a servo to pull it tight or relax it.
  • P!-RoP!-Ro Posts: 1,189
    edited 2009-11-08 05:57
    I'm actually from Blackfoot, Idaho, but I visit my family yearly in Oklahoma. My grandparents live in Norman so that is where I go usually, however I do have cousins in Moore. I can't say I'm the most knowledgeable of OU, however my grandfather used to teach there and their house is within walking distance of the school, so I have had the chance to see it many times.

    As for the locking mechanism, I was actually thinking about the possibility of a rubber pad pressed down on a plate to stop it from rotating, but that won't necessarily have too high of locking capabilities. Probably for the highest of strength and efficiency, a notched plate could be used, with electromagnetic lock able to slide a bar into the notches to hold position. This may also increase accuracy as the notches will hold more precise positions than something without limits, and possibly able to slide unintentionally as well.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    PG
  • ercoerco Posts: 20,256
    edited 2009-11-15 22:58
    So Dennis, are you planning to build one?

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    ·"If you build it, they will come."
Sign In or Register to comment.