robot_localization estimate (using only IMU) drifts for a stationary robot (Gazebo Model)

edit

I am trying to localize my robot (Gazebo model) in a known map. The robot has a lidar and IMU on it. To start with I tried visualizing the robot_localization estimate in rviz. I am feeding only IMU data to the ekf estimator. I did not expect the Odometry estimate to be very accurate, however even when the robot is not moving the estimate generated by robot_localization keeps on moving. Also the estimate covariance keeps on increasing (this can be observed in rviz). What could possibly be going wrong?

I am publishing a static transform to place the map frame at the odom frame. Here is how my tf tree looks like (only relevant part shown):

This is the robot_localization configuration:

frequency: 30
sensor_timeout: 0.05
two_d_mode: true
transform_time_offset: 0.0
transform_timeout: 1.0
print_diagnostics: false
debug: false
publish_tf: true
publish_acceleration: false

odom_frame: odom
base_link_frame: base_footprint
world_frame: odom

imu0: /imu

imu0_config: [false, false, false,
              false,  false,  true,
              false, false, false,
              false,  false,  true,
              true,  false,  false]

imu0_nodelay: false
imu0_differential: false
imu0_relative: true
imu0_queue_size: 5
imu0_pose_rejection_threshold: 0.8
imu0_twist_rejection_threshold: 0.8
imu0_linear_acceleration_rejection_threshold: 0.8

imu0_remove_gravitational_acceleration: true

dynamic_process_noise_covariance: true

process_noise_covariance: [ sample values from robot_localization github repo]

initial_estimate_covariance: [ value of 0.01 for variables set to true in imu0_config]

Here are the ROS IMU plugin settings:

<gazebo reference="base_imu_link">
    <gravity>true</gravity>
    <sensor name="razor_9dof_imu" type="imu">
        <always_on>true</always_on>
        <update_rate>100</update_rate>
        <topic>/imu</topic>
        <plugin filename="libgazebo_ros_imu_sensor.so" name="imu_plugin">
            <topicName>/imu</topicName>
            <bodyName>base_imu_link</bodyName>
            <updateRateHZ>100</updateRateHZ>
            <gaussianNoise>0.05</gaussianNoise>
            <xyzOffset>0 0 0</xyzOffset>
            <rpyOffset>0 0 0</rpyOffset>
            <frameName>base_imu_link</frameName>
        </plugin>
        <pose>0 0 0 0 0 0</pose>
    </sensor>
</gazebo>

Here is the IMU data:

header: 
  seq: 1791
  stamp: 
    secs: 37
    nsecs: 148000000
  frame_id: "base_imu_link"
orientation: 
  x: 0.0246637627368
  y: -0.026593164317
  z: 0.0668625611945
  w: 1.02289534517
orientation_covariance: [0.0025000000000000005, 0.0, 0.0, 0.0, 0.0025000000000000005, 0.0, 0.0, 0.0, 0.0025000000000000005]
angular_velocity: 
  x: -0.0326675653364
  y: -0.0340936938196
  z: -0.0323116318276
angular_velocity_covariance: [0.0025000000000000005, 0.0, 0.0, 0.0, 0.0025000000000000005, 0.0, 0.0, 0.0, 0.0025000000000000005]
linear_acceleration: 
  x: -0.0783367912198
  y: -0.00500801404049
  z: 9.79784343093
linear_acceleration_covariance: [0.0025000000000000005, 0.0, 0.0, 0.0, 0.0025000000000000005, 0.0, 0.0, 0.0, 0.0025000000000000005]