Setting covarince matrix for Robot Localization
Hi, I'm working with a custom made rover. I has 4 wheels, 2 of them with encoders. I have an onboard computer with an STM32F722 uC. I have an IMU and a Magnetometer connected to it. Usin rosserial node I publish sensor data to ROS master on a ODROID XU4. There I will run an imu_madgick_filter node in order to get orientation and then fusing it with odometry from wheel encoder. Now I want to set covariance matrices field values for each sensor. Each ROS message expect a covariance matrix for that sensor, in particular IMU and Odom. I would like to know a way to get this values. Can somenone point me out to some reading material for this subject? or can someone gives me some starting values?
EDIT1: I need to get a reliable orientation with imu_filter_madgwick as RayROS suggest. So I opened another question here.
EDIT2: I already test imu_filter_madgwick and it doesn't take into account covariance matrix from sensor data. It pass through whatever covariance the original message have to the output filtered imu data. So how can I get reliable values for angular_velocity_covariance and linear_acceleration_covariance on IMU messages from my sensor data
Thank