I've a robot that starting at a point moves certain distance and comes back to the start location.

I've used the methodology explained here to localize the robot as shown in the below picture. However, as you can see it is clearly error-prone in that the start and end positions are set apart while they are supposed to be the same point, which I ascribe to the error propagation intrinsic to the method.

The linked solution is prone to accumulated error and so would like to address it by leveraging this knowledge (of start and end points sharing the same coordinates, and opposite headings) through a feed-back controller to influence the determined trajectory so it starts following a rather more accurate looking path and the start and end points converge in effect.

I'm not sure if I can use any of the filtering tricks, but let's just say all I've is Left and Right wheel frequencies I used to trace what is supposed to be the trajectory followed by the robot.

enter image description here

Any help would be appreciated.

  • I don't have the time to go into any more of an answer at the moment, but I believe what you're looking for is a "loop closure" technique. As it looks like you're wanting to do this with odometry measurements, I'd suggestion you try search terms like "dead reckoning loop closure." Hopefully that will get you some papers that can give you methods that you're looking for. If you do find a solution (before someone else here posts an answer), please answer your own question below and accept that answer! Good luck :) – Chuck Nov 7 at 21:12

Your Answer


By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Browse other questions tagged or ask your own question.