tracker.yaml

# SPDX-License-Identifier: Apache-2.0
# Copyright (C) 2025 Intel Corporation
---
univloc_tracker:
  # ID of this tracker
  ID: 0

  # The name of world coodinate in tf tree. Default is map
  map_frame: 'map'

  # Whether to get camera info from topic
  get_camera_info_from_topic: true

  # Monocular, Stereo, RGBD, Monocular_Inertial, Stereo_Inertial or RGBD_Inertial
  camera_setup: 'RGBD'

  # Default prefix of all the input topics
  camera: 'camera'

  # Camera intrinsics, necessary if get_camera_info_from_topic equals false
  camera_intrinsics: [0.0]

  # Camera distortion, necessary if get_camera_info_from_topic equals false
  camera_distortion: [0.0]

  # FPS of input camera images, please change the value to align with the camera
  camera_fps: 30.0

  # Camera color order, options: "Gray", "RGB", "RGBA", "BGR", "BGRA"
  camera_color_order: 'RGB'

  # Topic of the color image; leave blank to use "$(arg camera)/color/image_raw"
  image_topic: ''

  # Topic of the image input of right channel for stereo image input
  right_image_topic: ''

  # Topic of the color image; leave blank to use "$(arg camera)/aligned_depth_to_color/image_raw"
  depth_topic: ''

  # Topic of camera info; leave blank to use (right_)image_topic with last substring replaced by "camera_info"
  camera_info_topic: ''

  # Topic of camera info of right channel in stereo image input;
  right_camera_info_topic: ''

  # If using raw transport when enabling image transport feature.
  raw_transport: true

  # mapping, localization or relocalization
  slam_mode: 'mapping'

  # The usage model of the camera, Perspective or Fisheye
  model_type: 'Perspective'

  # Size of the input data queue; set 1 in real-time usage to always process the latest image
  # Set 0 to specify an unlimited size to ensure every image be processed sequentially (with lag)
  queue_size: 1

  # Reset the system if consecutively lost for a specified number of frames; set 0 to never reset
  num_lost_frames_to_reset: 0

  # Vocabulary file, either filename under the config folder or an absolute path
  vocabulary: 'orb_vocab.dbow2'

  # Whether to publish the estimated pose onto /tf
  publish_tf: true

  # Child frame for publishing poses to /tf; usually "base_link" or "camera_link"
  # Leave blank to use image_frame
  pub_tf_child_frame: ''

  # Parent frame for publishing poses to /tf; usually "map"
  pub_tf_parent_frame: ''

  # The pose of the camera in the tf2 message
  image_frame: ''

  # The frame ID in tf tree that does not change over time
  tf_fix_frame: ''

  # Whether to get camera extrinsics (w.r.t. robot base) from /tf
  get_camera_extrin_from_tf: true

  # Timeout to query camera extrinsics from /tf
  camera_extrin_query_timeout: 10.0

  # Used only if get_camera_extrin_from_tf is true
  baselink_frame: 'base_link'

  # Whether to clean redundant keyframe
  clean_keyframe: true

  # Whether to fuse odom data
  use_odom: false

  # Frame name of odom, if use_odom is true
  odom_frame: 'odom'

  # Timeout in millisecond for querying odom pose
  # Will block image callback, should consider the frequency of image and odom
  odom_tf_query_timeout: 0.0

  # Will block tracking module, should consider the frequency of image and odom
  odom_buffer_query_timeout: 0.0

  ### ORB feature extraction parameters
  # Maximal number of keypoints
  max_num_keypoints: 1000

  # Image scaling factor of the pyramid
  orb_scale_factor: 1.2

  # Number of levels in the pyramid
  orb_num_levels: 8

  # Initial FAST threshold
  orb_ini_fast_threshold: 20

  # Minimum FAST threshold
  orb_min_fast_threshold: 7

  # Mask rectangle
  # Must provide 4 rectangles and each rectangle param is xmin, xmax, ymin, ymax;  Leave blank if no mask require
  # orb_mask_rectangles : [0.0, 1.0, 0.0, 0.1, 0.0, 1.0, 0.90, 1.0, 0.0, 0.2, 0.7, 1.0, 0.8, 1.0, 0.7, 1.0]
  orb_mask_rectangles: [0.0]

  # Mask Image
  # Provide mask image in PNG or JPEG file with region in black and white. Keypoints inside black region will be drop
  # Mask image width and height must match camera width and height
  orb_mask_image: ''

  ### More options
  # Whether to visualize keypoints in a window
  gui: true

  # Whether to launch rviz
  rviz: true

  # Whether to launch the tracker node with gdb for debugging
  gdb: false

  # The path to save the trajectory of tracker node
  traj_store_path: ''

  # Set spdlog log level: [trace, debug, info, warning, error, critical, off]
  log_level: 'info'

  # The transformation matrix between camera coordinate and image frame (T_ic)
  T_image_to_camera: [0.0]

  # For stereo cameras, it is the value of the baseline between the left and right cameras multiplied by the focal length fx.
  # If set as default value (0.0), it will automatically be set using camera_baseline and focal length fx.
  focal_x_baseline: 0.0

  # The distance between lenses of a stereo camera (unit in meter)
  # reference data for RealSense cameras: 50mm for D435, 55mm for D415, 95ms for D455
  camera_baseline: 0.05

  # Depth threshold factor for RGBD and stereo input
  # For stereo/RGBD input, it will be multiplied with camera_baseline to obtain the true depth threshold
  depth_threshold: 40.0

  # Scale factor of the depth values (in inverse meter; 1000 means 0.001 m)
  depthmap_factor: 1000.0

  # Whether to enable rectifier for stereo camera
  enable_rectifier: true

  # Whether stereo rectification parameters are provided
  # For public datasets, the recitifier params are given.
  # For self-recorded datasets, typically we need to obtain the recitifier params on our own.
  # Our system provides a way to obtain the params but currently the performance is not guaranteed.
  # Will need further investigation to verify.
  rectifier_params_given: true

  ### Stereo rectification parameters
  # Original intrinsic parameters for left camera
  K_left: [0.0]

  # Original distortion parameters for left camera
  D_left: [0.0]

  # Rotation matrix for left camera after stereo recitification
  R_left: [0.0]

  # Original intrinsic parameters for right camera
  K_right: [0.0]

  # Original distortion parameters for right camera
  D_right: [0.0]

  # Rotation matrix for right camera after stereo recitification
  R_right: [0.0]

  # The frame ID of right camera in tf tree for stereo camera
  tf_right_camera_frame: ''

  # The left camera to right camera coordinate when using stereo input
  left_to_right_transform: [0.0]

  # Frame name of imu, if camera_setup is Monocular_Inertial, Stereo_Inertial or RGBD_Inertial
  imu_frame: 'd400_imu'

  # Whether to get imu extrinsics (w.r.t. image frame) from /tf
  get_imu_extrin_from_tf: false

  # The transformation matrix between camera and imu, i.e. translation, quaternion (x, y, z, w, x, y, z)
  # if get_imu_extrin_from_tf is false
  tf_camera_imu: [0.020096886903, -0.00507855368778, -0.0115051260218, 0.999987, -0.000504782, 0.00444093, -0.00264019]

  # Either set imu_topic only, or set both accel_topic and gyro_topic
  # The topic of imu input, if accel_topic and gyro_topic are empty
  imu_topic: ''

  # The topic of the accelarator, if imu_topic is empty
  accel_topic: ''

  # The topic of the gyroscope, if imu_topic is empty
  gyro_topic: ''

  # The frequency parameter used to process the imu input
  # Referenced from RealSense_D435i.yaml in Examples/RGB-D-Inertial at ORBSLAM3 repo
  imu_frequency: 200.0

  # The bias parameter used to process the imu input, only used for initialization
  imu_bias: [0.0, 0.0, 0.0, 0.0, 0.0, 0.0]

  # The noise parameter used to process the imu input, i.e. ngx, ngy, ngz, nax, nay, naz
  # Referenced from RealSense_D435i.yaml in Examples/RGB-D-Inertial at ORBSLAM3 repo
  imu_noise: [1.0e-02, 1.0e-02, 1.0e-02, 1.0e-01, 1.0e-01, 1.0e-01]

  # The bias_walker parameter used to process the imu input, i.e. wgx, wgy, wgz, wax, way, waz
  # Referenced from RealSense_D435i.yaml in Examples/RGB-D-Inertial at ORBSLAM3 repo
  # The real numbers should be 1.0e-06, 1.0e-06, 1.0e-06, 1.0e-04, 1.0e-04, 1.0e-04
  # But the launch file cannot read those float numbers properly, thus using the below numbers
  imu_bias_walker: [1.0001e-06, 1.0001e-06, 1.0001e-06, 1.0e-04, 1.0e-04, 1.0e-04]

  # whether to force best effort QoS for ROS service
  force_best_effort: false

  # whether to use lidar or not
  use_lidar: false

  # Whether to get lidar extrinsics (w.r.t. robot base) from /tf
  get_lidar_extrin_from_tf: false

  # The transformation matrix between base_link to Lidar, i.e. translation, quaternion (x, y, z, w, x, y, z)
  # use tf_base_lidar if get_lidar_extrin_from_tf is false
  tf_base_lidar: [0.0, 0.0, 0.16, 0.0, 0.0, 0.0, 0.0]

  # Lidar displacement from the ground
  lidar_height: 0.22

  # Frame name of lidar
  lidar_frame: ''

  # namespace under which topics are created
  namespace: ''

  # Fast Mapping settings
  enable_fast_mapping: false

  # depth threshold for fastmapping, should be smaller than true depth threshold (camera_baseline * depth_threshold)
  depth_max_range: 3.3

  # Voxel size in meters
  voxel_size: 0.04

  # Minimum and maximum height, in meters, from which the camera input will be processed and integrated into the octree
  # If left empty, the entire camera input will be processed therefore the 3D map will contain floors and ceilings
  zmin: 0.30
  zmax: 0.60

  # Size of the volumetric map in voxels
  map_size: 256

  # Scaling factor of the depth image
  depth_scaling_factor: 2

  # The path to store octree map of tracker node
  octree_store_path: ''

  # The path to load pre-constructed octree map
  octree_load_path: ''

  # The four vertexes of remapping region, [P1.x, P1.y, ... , P4.x, P4.y], must be in clockwise or anti-clockwise
  # order
  remapping_region: [0.0]

  # Divisor to trigger relocalization in localization mode
  reloc_trig_divisor: 1

  # Compute and publish covariance matrix of pose in mapping mode, will further enable in localization and remapping mode
  need_covariance: false