Revision history [back]

Fundamentally, you need to get two parts right -

1. Localization - It is likely that your existing localization method (say AMCL) will not provide appropriate accuracy for this task. You need to use your LiDAR/Camera data to localize the robot relative to the table. This would involve some object detection algorithm to identify table's pose relative to the robot - ICP/RANSAC if you're using point cloud, an artificial marker like QR/AruCO or perhaps vision based detection based on deep learning.

2. Controls - Once you have the above figured out, you need a precise docking controller to reach the target with lots of safety in place. If you're using nav2/move_base then out of the box TEB/DWA/DWB will not solve your problem reliably.