I have an Inertial Navigation System (INS) that publishes position information in GPS and UTM formats. I have mounted the device on a robot. My objective is to move the robot from location A to location B and assess the accuracy of the INS in providing position information. I intend to input the initial position of the cart to the INS, move the robot precisely 20 meters, and then record the new position. Do you have any specific guidance for doing this test?
If this is not what you are looking for, then you need to specify your question and the context of your application more clearly.
You can use the robot_localization's implementation of the extended kalman filter to fuse all of the odometry data you have between camera, GPS, and IMU. This will give you the robot's relative position from it's starting point and you can literally send a move_base goal 20m ahead of it and manually check it's accuracy.
I would not recommend relying on visual or lidar odometry due to their inherent drift issues. Instead, I suggest employing Aruco markers. Create a large sheet with various Aruco markers of different sizes, ensuring they are visible at varying distances. By accurately estimating the robot's position relative to one of these markers, you can validate your data.