For our map, we borrowed a map from another robot and attempted to load into Stretch 3. Navigating and teleoping via RViz works well as there is a way to localize the robot, but I was having trouble localizing/navigating the robot on web teleop. Any suggestions for using a borrowed map for Stretch via web teleop?
navigation map png (later converted to pgm but since the forum here does not support pgm, I have uploaded png. Also note that when testing, I have updated the map name to nav2_demo_map.pgm) and yaml:
Hi @zyaskawa, good question. The web interface doesn’t have a GUI for setting the robot’s location. This is something we’d need to develop. I recently wrote a tutorial, Developing on the Web Interface, that covers React, ROS2, and adding new GUI to the interface, so if this is something you’re interested in trying to implement, I’d recommend reading the tutorial. @hello-vinitha, the project lead on the interface, might also be interested.
Alternatively, if you know where the robot wakes up in the map (i.e. its initial pose), you can hard-code it into a node that publishes to the /initialpose topic when the launch file is called. Even if the robot isn’t exactly at the hard-coded pose, if it starts close to the hard-coded estimate, it will converge on the true pose as the robot starts moving. This is due to the “motion update” of AMCL, a particle filter that estimates the pose of the robot. I have a post that describes how to call /request_nomotion_update to cause AMCL to converge, even when the robot isn’t moving.
