SLAM doesn't work properly and mapping jumps between points

[aliwaqas333]

[ERROR] [1627181523.437889922]: Transform failed: "odom" passed to lookupTransform argument source_frame does not exist.

i am trying to run openvslam with Intel realsense D435 on my msi laptop. I tried to comment the transform, i-e line 47 from openvslam_ros.cc , this made the error go away but still the SLAM doesn't work properly and mapping jumps between points.

    if (publish_tf_) {

Please help in this regards.

Please check this video to understand the problem.

Video

Video of the issue with screen recording

Launch File

<launch>
    <!-- open realsense RGBD mode -->
    <include file="$(find realsense2_camera)/launch/rs_rgbd.launch" />

    <!-- open openVSLAM node  -->
    <node name="openvslam_node" pkg="openvslam_ros" type="run_slam" 
    args="-v $(find openvslam_ros)/orb_vocab/orb_vocab.fbow -c $(find openvslam_ros)/camera_config/rs_D435i_rgbd.yaml --map-db $(find openvslam_ros)/maps/map8.msg"
    output="screen"/>

    <!-- rename depth topics from realsense  -->
    <node name="image_transport_node" pkg="image_transport" type="republish" 
    args="raw in:=/camera/aligned_depth_to_color/image_raw raw out:=/camera/depth/image_raw"/>


</launch>

Yaml File

My yaml config file is:

Camera:
  name: "RS_D435 RGBD"
  setup: "RGBD"
  model: "perspective"

  fx: 608.6885681152344
  fy: 608.63482666015625
  cx: 311.8787841796875
  cy: 247.81253051757812

  k1: 0
  k2: 0
  p1: 0
  p2: 0
  k3: 0

  fps: 30.0
  cols: 640
  rows: 480

  # baseline is 50 mm
  focal_x_baseline: 36.3554139709

  color_order: "RGB"

#=====================#
# Tracking Parameters #
#=====================#

depth_threshold: 40.0
depthmap_factor: 1000.0
# ref: https://github.com/IntelRealSense/realsense-ros/issues/41

#================#
Feature:
  max_num_keypoints: 2000
  scale_factor: 1.2
  num_levels: 8
  ini_fast_threshold: 20
  min_fast_threshold: 7

#====================#
# Mapping Parameters #
#====================#

Mapping:
  baseline_dist_thr_ratio: 0.02

Initializer:
num_min_triangulated_pts: 400
#===========================#
# PangolinViewer Parameters #
#===========================#

PangolinViewer:
  keyframe_size: 0.06
  keyframe_line_width: 1
  graph_line_width: 1
  point_size: 2
  camera_size: 0.07
  camera_line_width: 2
  viewpoint_x: 0
  viewpoint_y: -20
  viewpoint_z: -0.1
  viewpoint_f: 2800

Thanks for taking time to read this.
#37

[ymd-stella]

I checked the video and the camera parameters (fx, fy, k1, etc.) shown seem to be different from what you have posted here.

[aliwaqas333]

I checked the video and the camera parameters (fx, fy, k1, etc.) shown seem to be different from what you have posted here.

Hi, I have been changing the values to see if it fixes the issue. Do you think it could be a problem with parameters?

[ymd-stella]

is it possible to provide map x, y, z of robot to slam to remove errors. my robot gets its exact location from time to time by recognizing some special markers. Can i provide that location and pose data to slam.

What is the purpose?

Can you please tell me what will be benefit of providing tf from map to odom?

See https://www.ros.org/reps/rep-0105.html

[aliwaqas333]

the purpose is to make slam more accurate.

Also is there any function to reset the slam? Like there is a gui button but is there any ros service? thanks

[aliwaqas333]

about the baselink, is it some matrix? what is the shape and how to pass it to openvslam?

[ymd-stella]

the purpose is to make slam more accurate.

Why do you use Visual SLAM? If markers are more accurate than Visual SLAM, can't they be estimated by markers only?

Also is there any function to reset the slam? Like there is a gui button but is there any ros service?

No. Pull requests are welcome.

about the baselink, is it some matrix? what is the shape and how to pass it to openvslam?

tf2_msgs/msg/TFMessage. You probably don't know much about tf2. It is highly recommended that you learn how to use tf2 well first.

[aliwaqas333]

Markers are accurate because they were already placed there and location values are encoded in them. but there are very few markers and i want to to fuse them to make slam more accurate. thanks.

Thank you so much. I am learning and will definitely contribute to this community.

[nitishsaDire]

[ERROR] [1627181523.437889922]: Transform failed: "odom" passed to lookupTransform argument source_frame does not exist.

i am trying to run openvslam with Intel realsense D435 on my msi laptop. I tried to comment the transform, i-e line 47 from openvslam_ros.cc , this made the error go away but still the SLAM doesn't work properly and mapping jumps between points.

    if (publish_tf_) {

Please help in this regards.

Please check this video to understand the problem.

Video

Video of the issue with screen recording

Launch File

<launch>
    <!-- open realsense RGBD mode -->
    <include file="$(find realsense2_camera)/launch/rs_rgbd.launch" />

    <!-- open openVSLAM node  -->
    <node name="openvslam_node" pkg="openvslam_ros" type="run_slam" 
    args="-v $(find openvslam_ros)/orb_vocab/orb_vocab.fbow -c $(find openvslam_ros)/camera_config/rs_D435i_rgbd.yaml --map-db $(find openvslam_ros)/maps/map8.msg"
    output="screen"/>

    <!-- rename depth topics from realsense  -->
    <node name="image_transport_node" pkg="image_transport" type="republish" 
    args="raw in:=/camera/aligned_depth_to_color/image_raw raw out:=/camera/depth/image_raw"/>


</launch>

Yaml File

My yaml config file is:

Camera:
  name: "RS_D435 RGBD"
  setup: "RGBD"
  model: "perspective"

  fx: 608.6885681152344
  fy: 608.63482666015625
  cx: 311.8787841796875
  cy: 247.81253051757812

  k1: 0
  k2: 0
  p1: 0
  p2: 0
  k3: 0

  fps: 30.0
  cols: 640
  rows: 480

  # baseline is 50 mm
  focal_x_baseline: 36.3554139709

  color_order: "RGB"

#=====================#
# Tracking Parameters #
#=====================#

depth_threshold: 40.0
depthmap_factor: 1000.0
# ref: https://github.com/IntelRealSense/realsense-ros/issues/41

#================#
Feature:
  max_num_keypoints: 2000
  scale_factor: 1.2
  num_levels: 8
  ini_fast_threshold: 20
  min_fast_threshold: 7

#====================#
# Mapping Parameters #
#====================#

Mapping:
  baseline_dist_thr_ratio: 0.02

Initializer:
num_min_triangulated_pts: 400
#===========================#
# PangolinViewer Parameters #
#===========================#

PangolinViewer:
  keyframe_size: 0.06
  keyframe_line_width: 1
  graph_line_width: 1
  point_size: 2
  camera_size: 0.07
  camera_line_width: 2
  viewpoint_x: 0
  viewpoint_y: -20
  viewpoint_z: -0.1
  viewpoint_f: 2800

Thanks for taking time to read this. #37

Could you please explain how you got config.yaml file for your realsense camera?