Home » Building, Ground robots, Ideas, Issues, Refining the project » Improving the omnidirectional robot design

Improving the omnidirectional robot design

It’s been two weeks now we have been struggling with a major issue. After implementing the manual driving mode, we wanted the robot to move to specific positions and we encountered problems with accuracy. For instance, on a same command, the robot could have different behaviors, totally unpredictable. We sure knew this could be improved thanks to several corrections (like a PID that we’ve been working on recently) but still, we thought those errors couldn’t be only due to the mechanics. Finally, we found that our error was tremendously amplified by the orientation of the omniwheels.


Indeed, instead of putting the omniwheels with their axis parallel to the floor, we decided to use some angle for stability and design purposes. But with some hindsight, we just understood that the omniwheels are just meant to be used completely perpendicular to the floor for some physical properties. When you take a normal wheel, whatever angle you put it on the floor, the only translation you can do is along the line parallel to the floor and perpendicular to the axis of the wheel. Setting the omniwheel with a none right angle messed the previous rule just mentioned.



And this is the reason why we came out with a new design for our ground units that you can see just above. Maybe you just noticed that the omniwheels are different too, and you would be right if you did. We had to change it to because the old design didn’t allow a perpendicular setup. If you’re interested in this new design, it’s one of our own that you can use and we’re delivering some pictures below if you need a pattern.



Now we’re working on a new system of movement: acknowledging the robot is located at the point (0;0), he’s supposed to move to any specific (x;y) position. The system is working according to our expectations but we need to work on the correction to make it more reliable (because we’re loosing significant accuracy within time and distance). We’ll try to show some videos with and without corrections on the next post.

One Response to Improving the omnidirectional robot design

  1. An extreme example of a tilted omniwheel (10 degrees) can be seen at http://www.kornylak.com/wheels/transdisc.html a video of a powered application can be acessed at the bottom of the page.
    Tom Kornylak