I am working on a SLAM build for the turtlebot. When I go to the laboratory setting it appears that each time I must redo my mapping and localization before I try navigating. Does anyone know why this may the case and how I could avoid it? It takes a lot of time to do mapping and localization, and I don’t have too much time left to do actual tests. Note that when I am working with the simulation, I only have to do the mapping and localization once.
You should be able to use any map and amcl node with an initial position.
What are you doing differently? You can even use the same map you used in simulation (although it might not be as precise) and then just set the initial position in rviz based on where the robot is
I am not fully certain what might be the error. I have observed on a number of occasions that if I perform localization without doing mapping first, then the points I record do not appear to reflect the points on the saved map. Going forward I will need to double check to ensure I am setting the initial points in rviz each time I do the localization. It’s almost as if each time I go into the lab the map is assigned a new coordinate system. So the points I record via localization are no longer valid and I need to redo the process again. I will double check to make sure this isn’t an issue related to assigning initial points. Any other potentially valid ideas are welcomed as well.
I thought you meant the initial pose of the amcl node, not the points you then obtain in your own application.
How do you capture these points? If it’s with odometry, then it will definitely be different in the real robot than the simulation.
You can introspect the differences by looking for example at the frames in rviz, and look for differences that would explain the behavior. I don’t think anything is wrong here, it’s just the switching from one environment to another.
It is using a similar methodology to the standard SLAM approach discussed in one of the ROS courses taught on the subject. I cannot recall exactly what it is using, though I suspect it is using odometry. I think the main issue had to do with a failure to assign an initial position before doing tasks related to localization or mapping. I will need to better document the issue next time. I wanted to see if there was any common issues related to using SLAM in the lab that people may have encountered and were inline with what I was observing, but did not appear to be related to initial points.