noetic

Nodes

Visual and Lidar Odometry

Common odometry stuff for rgbd_odometry, stereo_odometry and icp_odometry nodes. These nodes wrap the various odometry approaches of RTAB-Map. When a transformation cannot be computed, a null transformation is sent to notify the receiver that odometry is not updated or lost.

• There are two set of parameters: ROS and RTAB-Map's parameters. The ROS parameters are for connection stuff to interface the RTAB-Map library with ROS. The RTAB-Map's parameters are those from the RTAB-Map library.
  • Parameters available for odometry can be shown from the terminal by using the "--params" argument:

    • $ rosrun rtabmap_odom rgbd_odometry --params
      or
      $ rosrun rtabmap_odom stereo_odometry --params
      or
      $ rosrun rtabmap_odom icp_odometry --params

Arguments

• "--params": Show RTAB-Map's parameters related to this node and exit.

Subscribed Topics

imu (sensor_msgs/Imu)

• IMU used by VIO approaches (see rtabmap --params | grep Odom/Strategy). For non-VIO approaches, it can be used to initialize the odometry to be aligned with gravity (with wait_imu_to_init=true). For default visual F2M approach, gravity constraints will be added in local bundle adjustment, making it as a loosely VIO approach. For non-VIO approaches and F2M, the IMU topic should contain the quarternion already estimated (see imu_filter_madgwick or imu_complementary_filter).

Published Topics

odom (nav_msgs/Odometry)

• Odometry stream. Send null odometry if cannot be computed.

odom_info (rtabmap_msgs/OdomInfo)

• Odometry info stream. RTAB-Map's parameter Odom/FillInfoData should be true to fill features information (default is true).

odom_last_frame (sensor_msgs/PointCloud2)

• 3D features of the last frame used to estimate the transformation.

odom_local_map (sensor_msgs/PointCloud2)

• Local map of 3D features used as reference to estimate the transformation. Only published if RTAB-Map's parameter Odom/Strategy=0.

Services

reset_odom (std_srvs/Empty)

• Restart odometry to identity transformation.

reset_odom_to_pose (rtabmap_msgs/ResetPose)

• Restart odometry to specified transformation. Format: "x y z roll pitch yaw".

pause_odom (std_srvs/Empty)

• Pause odometry.

resume_odom (std_srvs/Empty)

• Resume odometry.

Parameters

~frame_id (string, default: "base_link")

• The frame attached to the mobile base.

~odom_frame_id (string, default: "odom")

• The odometry frame.

~publish_tf (bool, default: "true")

• Publish TF from /odom to /base_link.

~tf_prefix (bool, default: "")

• Prefix to add to generated tf.

~wait_for_transform (bool, default: "false")

• Wait (maximum 1 sec) for transform when a tf transform is not still available.

~initial_pose (string, default: "")

• The initial pose of the odometry. Format: "x y z roll pitch yaw".

~queue_size (int, default: 10)

• Size of message queue for each synchronized topic.

~publish_null_when_lost (bool, default: true)

• Odometry is published with null transform when lost.

~ground_truth_frame_id (string, default: "")

• This can be used to have odometry initialized at the current ground truth pose.

~ground_truth_base_frame_id (string, default: "")

• See ground_truth_frame_id description.

~guess_frame_id (string, default: "")

• Use this frame as guess to compute odometry, otherwise odometry guess is done from a constant motion model. tf published by this node will be the correction of actual odometry. If guess_frame_id is odom_combined, odom_frame_id is odom and frame_id is base_footprint, odom -> odom_combined is published by this node to have a tf tree like this: /odom -> /odom_combined -> /base_link.

~guess_min_translation (float, default: 0.0 m)

• When guess_frame_id is set, don't update odometry unless the guess moved at least as much as this value.

~guess_min_rotation (float, default: 0.0 rad)

• When guess_frame_id is set, don't update odometry unless the guess rotated at least as much as this value.

~config_path (string, default: "")

• A config (ini) file with RTAB-Map's parameters.

~wait_imu_to_init (bool, default: false)

• Make the node waits for first imu before starting estimating odometry. This will make sure that odometry is aligned with gravity on initialization.

Required tf Transforms

base_link → <the frame attached to sensors of incoming data>

• usually a fixed value, broadcast periodically by a robot_state_publisher, or a tf static_transform_publisher.

Provided tf Transforms

odom → base_link

• the current estimate of the robot's pose within the odometry frame.

rgbd_odometry

This node wraps the RGBD odometry approach of RTAB-Map. Using RGBD images, odometry is computed using visual features extracted from the RGB images with their depth information from the depth images. Using the feature correspondences between the images, a RANSAC approach computes the transformation between the consecutive images.

See also Visual Odometry for common odometry stuff used by this node.

Subscribed Topics

rgb/image (sensor_msgs/Image)

• RGB/Mono rectified image.

rgb/camera_info (sensor_msgs/CameraInfo)

• RGB camera metadata.

depth/image (sensor_msgs/Image)

• Registered depth image.

rgbd_image (rtabmap_msgs/RGBDImage)

• RGB-D synchronized image, only when subscribe_rgbd is true and rgbd_cameras=1.

rgbd_image# (rtabmap_msgs/RGBDImage)

• RGB-D synchronized image, only when subscribe_rgbd is true and rgbd_cameras>=2. Index # starting at 0.

rgbd_images (rtabmap_msgs/RGBDImages)

• Multi synchronized cameras, only when subscribe_rgbd is true and rgbd_cameras=0

Parameters

~approx_sync (bool, default: "true")

• Use approximate time synchronization of rgb and depth messages. If false, the rgb and depth images must have the same timestamps.

~subscribe_rgbd (bool, default: "false")

• Subsribe to rgbd_image topic.

~rgbd_cameras (int, default: 1)

• Number of RGB-D cameras to use (when subscribe_rgbd is true). Well for now, a maximum of 4 cameras can be synchronized at the same time. If > 1, the rgbd_image topics should contain the camera index starting with 0. For example, if we have 2 cameras, you should set rgbd_image0 and rgbd_image1 topics. If set to 0, rgbd_images (published by a rgbdx_sync node) input is used instead, internal images should be RGB-D, not stereo. If they are stereo, use stereo_odometry node instead.

stereo_odometry

This node wraps the stereo odometry approach of RTAB-Map. Using stereo images, odometry is computed using visual features extracted from the left images with their depth information computed by finding the same features on the right images. Using the feature correspondences, a RANSAC approach computes the transformation between the consecutive left images.

See also Visual Odometry for common odometry stuff used by this node.

Subscribed Topics

left/image_rect (sensor_msgs/Image)

• Left RGB/Mono rectified image.

left/camera_info (sensor_msgs/CameraInfo)

• Left camera metadata.

right/image_rect (sensor_msgs/Image)

• Right Mono rectified image.

right/camera_info (sensor_msgs/CameraInfo)

• Right camera metadata.

rgbd_image (rtabmap_msgs/RGBDImage)

• Stereo synchronized image, only when subscribe_rgbd is true and rgbd_cameras=1. This topic should contain left image in rgb field and right image in depth field (with corresponding camera info).

rgbd_image# (rtabmap_msgs/RGBDImage)

• Stereo synchronized image, only when subscribe_rgbd is true and rgbd_cameras>=2. Index # starting at 0. This topic should contain left image in rgb field and right image in depth field (with corresponding camera info).

rgbd_images (rtabmap_msgs/RGBDImages)

• Multi synchronized cameras, only when subscribe_rgbd is true and rgbd_cameras=0

Parameters

~approx_sync (bool, default: "false")

• Use approximate time synchronization of stereo messages. If false, the left/right images and the left/right camera infos must have the same timestamps.

~subscribe_rgbd (bool, default: "false")

• Subsribe to rgbd_image topic.

~rgbd_cameras (int, default: 1)

• Number of stereo cameras to use (when subscribe_rgbd is true). Well for now, a maximum of 4 cameras can be synchronized at the same time. If > 1, the rgbd_image topics should contain the camera index starting with 0. For example, if we have 2 cameras, you should set rgbd_image0 and rgbd_image1 topics. If set to 0, rgbd_images (published by a rgbdx_sync node) input is used instead, internal images should be stereo, not RGB-D. If they are RGB-D, use rgbd_odometry node instead.

icp_odometry

This node wraps the icp odometry approach of RTAB-Map. Using laser scan or a point cloud, odometry is computed by Iterative Closest Point (ICP) registration.

See also Visual Odometry for common odometry stuff used by this node.

Subscribed Topics

scan (sensor_msgs/LaserScan)

• Laser scan, assuming the lidar is installed horizontally. Don't use scan or scan_cloud at the same time.

scan_cloud (sensor_msgs/PointCloud2)

• Point cloud, which can be 2D or 3D. Don't use scan or scan_cloud at the same time.

Parameters

~scan_cloud_max_points (int, default: 0)

• Maximum point in laser scan or point cloud. 0 means the maximum is defined by the size of ranges vector for laser scan topic.

~scan_downsampling_step (int, default: 1)

• Downsample the laser scan or point cloud. <=1 means no downsampling, otherwise one ray/point each step is kept.

~scan_voxel_size (float, default: 0.0 m)

• Filter the laser scan or point cloud using voxel filter of this size. 0.0 means disabled.

~scan_normal_k (int, default: 0)

• Estimate normals of the laser scan or point cloud by using k nearest points. Useful only if Icp/Point2Plane parameter is enabled. 0 means disabled.

~scan_normal_radius (float, default: 0.0)

• Estimate normals of the laser scan or point cloud by using nearest points under this fixed radius. Useful only if Icp/Point2Plane parameter is enabled. 0.0 means disabled.

~deskewing (bool, default: false)

• For PointCloud2 input scan_cloud (3D LiDAR), each point will be deskewed based on their timestamp during the rotation of the LiDAR by either using guess_frame_id when provided or using a constant velocity model. The PointCloud2 topic should have timestamp channel as t, time or stamps. Supported timestamp formats are FLOAT32 (seconds since header's timestamp) and UINT32 (nanoseconds since header's timestamp). If PointCloud2 input is organized, deskewing will be faster (transformed per columns instead per points). Note that for 2D LaserScan topic scan, it is deskewed only when guess_frame_id is used or frame_id is not the same frame than in the message.

~deskewing_slerp (bool, default: false)

• When enable, a lookup transform for TF will only be done for the first and last stamp, then points with in between timestamps will be deskewed by interpolating between the oldest and latest poses. This can be faster than the default approach looking for transform for very timestamps (though later approach would be more accurate).