Quote:
Originally Posted by AustinSchuh
As with all engineering, you need to figure out how much is good enough, and how to deal with things not being exactly as planned. Say someone gets between the wall and your sensor? Say someone crashes into your side? Do you care? We chose to just use the ground speed encoders and let someone crashing into our side make us loose track of our real position. For the most part, it worked fine, and it's debatable whether having that information when someone did hit us would have allowed us to do much more. Assuming you tried to add in an Ultrasonic sensor, that would only let you narrow down the potential error in one direction. I'd say just integrate the ground speed encoders, or the ones on your drive train, and use that until you have determined that that isn't enough. Detecting position and using it are isolated things that you can improve independently.
|
Considering that we will probably only have 15 seconds again, encoder errors shouldn't compound too quickly. Only if we go longer then that, a recalibration may be necessary
Quote:
|
If you are trying to point the robot in a direction in order to go in that direction, the error is some measurement that is fairly linear and will be 0 when you are pointed in the correct direction. The heading error was what I used when trying to go to a way point.
|
so you took your spline and split it into discrete points? then calculated the angle to the next point, and arced to the point, or did you turn, then drive straight?