Robot localization for very accurate position [closed]
##Scenario I have to do lane following in an environment where lanes are marked on the ground and many obstacles are present in those lanes. At some point, I must leave the lanes and navigate through another complex open area to reach another section of lanes. The course I am navigating is complex and requires a very accurate (less than 1m) mapping.
I already can identify the edges of lanes (including distance and shape) pretty well and I have a Lidar for the obstacles. Additionally, I have encoder, GPS, and IMU values.
##Question My question is this: If I treat the lane markings like obstacles (lines) I would receive from the Lidar and throw them into robot_localization with GPS and/or IMU data, then will I get an accurate position even if that position is only relative to my start point?
robot_localization provides IMU and odometry-based state estimation with a kalman filter, but it sounds like you're looking for something more like SLAM (Simultaneous Localization and Mapping.)
Seconded. r_l has no real concept of objects and obstacles. It just fuses pose, velocity, and acceleration data to provide a state estimate.
Can this question be closed, or are you still looking for more information?