Hello everybody,

We tried to setup the robot_localization and navigation2 stack on our robot running ROS2 Foxy.

Our robot works with the following sensor inputs:

• Wheel odometry (topic /odom)
• RTK GPS F9P (topic gps_position) for GPS position
• RTK GPS F9H (topic gps_heading) for GPS heading, connected to the F9P in a moving base setup

We are not using AMCL for localization because our robot has no LIDAR. We are not using any global costmap either because we do not need obstacle avoidance (global costmap set with large value of height and width in a rolling_window setup).

We are using the dual ekf configuration of robot localization, but using the gpsheading to replace the orientation data usually provided by an IMU in the robotlocalization package.

However, we have two main problems:

1. First problem is that we can observe a jagged trajectory for the odometry/filtered/global from the global ekf of robotlocalization. We are unsure about how we should interpret this trajectory (see attached ekflocalization.png file). Is it the expected result of robot_localization in a dual ekf setup or are we missing something ?

2. Second problem is that when we try to use the waypoint follower of navigation2 and robotlocalization is running using above setup, the robot is stuck in a weird state where it can only run a spinrunning recovery behavior despite the path planning working well. Do you have any advice for us ?

Please see attached our localization.yaml file for robot_localization, a schematic of our localization setup, our navigation.yaml file for navigation2. (EDIT: sorry for the attachements, I actually have not enough karma)

Thank you in advance for your help