AMCL Tuning Suggestions?

Hi everyone, I am trying to implement AMCL on a rosbag. The map used is generated from the same bag. I haven't used AMCL a lot and so I am unsure as to what parameters do what. It is appearing like my laser scan drifts with linear motion, I thought AMCl was suppose to compare the laser scan to the map and then make a probabilistic decision as to where it was? I have attached a video of what is occurring to try and tune this. [https://youtu.be/0AKt0TsxXjU](https://youtu.be/0AKt0TsxXjU) Thanks in advance for all assistance provided. PS. the AMCL params loaded is empty cause I do not know what I should tune. - **EDIT** I have spent some time tuning it and have achieved this behaviour, I have read that my choice in making the odom_alpha values larger causes a higher computational load, I am still interested in suggestions to make the performance better. https://youtu.be/edF884Wh4tA https://youtu.be/qQMIpvoGSWo laser_z_hit: 0.8 laser_z_rand: 0.2 odom_alpha1: 0.8 odom_alpha2: 0.4 odom_alpha3: 8 odom_alpha4: 0.6 launch: ekf param: ekf_se_odom: frequency: 10 sensor_timeout: 0.5 two_d_mode: true transform_time_offset: 0.0 transform_timeout: 0.0 print_diagnostics: true debug: false debug_out_file: /path/to/debug/file.txt publish_tf: true publish_acceleration: false map_frame: map # Defaults to "map" if unspecified odom_frame: odom # Defaults to "odom" if unspecified base_link_frame: base_link # Defaults to "base_link" if unspecified world_frame: odom # Defaults to the value of odom_frame if unspecified odom0: /nightrider/odom odom0_config: [false, false, false, false, false, false, true, false, false, false, false, true, false, false, false] odom0_queue_size: 5 odom0_nodelay: false odom0_differential: false odom0_relative: false odom0_pose_rejection_threshold: 5 odom0_twist_rejection_threshold: 1 imu0: /tx2/imu_enu imu0_config: [false, false, false, false, false, false, false, false, false, false, false, true, true, false, false] imu0_nodelay: false imu0_differential: false imu0_relative: true imu0_queue_size: 5 imu0_remove_gravitational_acceleration: true # imu0_pose_rejection_threshold: 0.8 # Note the difference in parameter names # imu0_twist_rejection_threshold: 0.8 # # imu0_linear_acceleration_rejection_threshold: 0.8 # use_control: false # [ADVANCED] The process noise covariance matrix can be difficult to tune, and can vary for each application, so it is # exposed as a configuration parameter. This matrix represents the noise we add to the total error after each # prediction step. The better the omnidirectional motion model matches your system, the smaller these values can be. # However, if users find that a given variable is slow to converge, one approach is to increase the # process_noise_covariance diagonal value for the variable in question, which will cause the filter's predicted error # to be larger, which will cause the filter to trust the incoming measurement more during correction. The values are # ordered as x, y, z, roll, pitch, yaw, vx, vy, vz, vroll, vpitch, vyaw, ax, ay, az. Defaults to the matrix below if # unspecified. process_noise_covariance: [1e-3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1e-3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1e-3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0.3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0.3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0.01, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0.5, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0.5, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0.1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0.3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0.3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0.3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0.3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0.3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0.3] # [ADVANCED] This represents the initial value for the state estimate error covariance matrix. Setting a diagonal # value (variance) to a large value will result in rapid convergence for initial measurements of the variable in # question. Users should take care not to use large values for variables that will not be measured directly. The values # are ordered as x, y, z, roll, pitch, yaw, vx, vy, vz, vroll, vpitch, vyaw, ax, ay, az. Defaults to the matrix below #if unspecified. initial_estimate_covariance: [1e-9, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1e-9, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1e-9, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1.0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1.0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1e-9, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1.0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1.0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1.0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1.0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1.0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1.0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1.0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1.0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1.0]