| ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange |
I am trying to visualize a custom osm file inside rviz using Autoware version 1.0.0 ade. The avp demo ros2 launch autoware_auto_avp_demo ms3_sim.launch.py runs without issues but I am not familiar enough with the program to change the latitude and longitude to match the same coordinates as my osm file.
Any advice on this?
Autoware.Auto uses the earth->map transform (usually published by the point cloud map loader) to then project the Lanelet2 map to have the same origin. In other words, when you provide an origin for your point cloud map (PCD file) with the YAML file that you provide to the point cloud map loader, it uses the lat/lon origin of the PCD to project your Lanelet2 map to have the same origin, so the YAML file is what you need to change.
If you are not using the point cloud map publisher, you must publish an earth->map transform yourself using something like a static_transform_publisher.
Hi Josh, I had implemented a GNSS+IMU based localization system for my purpose as I don't have LiDAR sensor on my robot. While I was testing the system on AutnomouStuff map, I find that there is some rotational variation between the point I get in map frame vs the point it was when we have PCD. The static transform publisher between earth->map I am publishing has only translation (this I had taken from Autoware.Auto demo I ran for AVP using NDT). Is there a rotation component as well for the map relative to earth that was not published in the static transform? I was able to find the rotation required (using the initial pose given for ndt_initializer) but it would be in map frame after translation and I don't know how to convert that to the rotation relative to earth frame, I am trying to calculate ...
I had debugged the issue a bit more and I find that the reason is most problably the GNSS data that is being published by the SVL simulator, it is not accurate. I checked it with few points on the map and saw that there is a need for both a translation component and a rotation component to align those same points with the values that I obtained for /earth->/base_link by calculation in the NDT matching demo. Basically, there is some error in lat, long, alt values returned by the SVL simulator and that error is probably not constant across the map (needs to be checked more) and this is causing the transforms to go wrong as we use /earth->/base_link transform in calculating /map->/odom here (unlike in NDT case where we use /map->/base_link obtained from NDT)
Asked: 2021-08-16 12:25:22 -0500
Seen: 207 times
Last updated: Sep 12 '21
ROS Answers is licensed under Creative Commons Attribution 3.0 Content on this site is licensed under a Creative Commons Attribution Share Alike 3.0 license.
