Revision history [back]

As NEngelhard says, 1mm is hard to achieve. Check out this paper, where the authors try to achieve maximum accuracy with laser scanners.

Accurate localization requires a precise map, so maybe try an approach as described in this paper

Stereo vision has measurement noise that grows quadratically with distance, therefore you won't get that good accuracy in general.

Laser scanners have different inaccuracies. For example, from what I heard from users, SICK scanners are more accurate than Hokuyo, but this might be model dependent.

Fusion of different sensors usually improves results. But also using more than one of the same type may be advantageous. The guys in the first paper use two laser scanners, with opposite viewing direction, which helps a lot.

As NEngelhard says, 1mm is hard to achieve. Check out this paper, where the authors try to achieve maximum accuracy with laser scanners.

Accurate localization requires a precise map, so maybe try an approach as described in this paper

Stereo vision has measurement noise that grows quadratically with distance, therefore you won't get that good accuracy unless the distance to obstacles is always in general.a specific range.

Laser scanners have different inaccuracies. For example, from what I heard from users, SICK scanners are more accurate than Hokuyo, but this might be model dependent.

Fusion of different sensors usually improves results. But also using more than one of the same type may be advantageous. The guys in the first paper use two laser scanners, with opposite viewing direction, which helps a lot.