rtabmap
rtabmap copied to clipboard
introlab
enhancing mapping kinect v1 ros 1 noetic and move base navigation
hello hope you're doing well so am using a kinect v1 only based slam and am kinda having not quite the results that i want when i map how can i improve it make it smoother and more accurate and also i want it to map the floor quickly so what params that can help in that way , also how should be my move_base files and what move base methods should i use for the navigation i want to be guided because am only using kinect so i kinda got stuck and thank you here some exemple that i got and also here is my launch file :
<launch>
<!-- ================== ARGS ================== -->
<arg name="model" default="dd_robot6.urdf"/>
<arg name="use_rtabmapviz" default="true"/>
<arg name="use_vo" default="true"/> <!-- visual odom node -->
<arg name="use_ekf" default="true"/>
<arg name="virtual_scan" default="true"/>
<arg name="rviz_config" default="$(find robot_diff)/urdf.rviz"/>
<arg name="db_path" default="$(env HOME)/.ros/rtabmap.db"/>
<!-- ================== ROBOT DESCRIPTION ================== -->
<param name="robot_description" textfile="$(find robot_diff)/urdf/$(arg model)"/>
<node name="joint_state_publisher" pkg="joint_state_publisher" type="joint_state_publisher"/>
<node name="robot_state_publisher" pkg="robot_state_publisher" type="robot_state_publisher"/>
<!-- ===== STATIC TF: base_link -> camera_link (EDIT the 6 numbers!) ===== -->
<node pkg="tf2_ros" type="static_transform_publisher" name="cam_tf"
args="0.2 0 0.1 0 0 0 base_link camera_link"/>
<node pkg="topic_tools" type="relay" name="grid_map_relay" output="screen" args="/rtabmap/grid_map /map" />
<!-- ===== diff_tf disabled (don't publish odom->base_link) -->
<node pkg="robot_diff" type="diff_tf.py" name="diff_tf" output="screen">
<param name="publish_tf" value="false"/>
</node>
<!-- ===== Trajectory viz ===== -->
<node name="robot_trajectory_visualization"
pkg="robot_diff"
type="trajectory_visualization.py"
output="screen"/>
<!-- ===== Rosbridge ===== -->
<include file="$(find rosbridge_server)/launch/rosbridge_websocket.launch">
<arg name="port" value="9090"/>
</include>
<!-- ===== RViz ===== -->
<node name="rviz" pkg="rviz" type="rviz"
args="-d $(arg rviz_config)"
required="true" />
<param name="use_sim_time" value="false"/>
<!-- ================== VISUAL ODOMETRY (rgbd_odometry) ================== -->
<group if="$(arg use_vo)">
<node pkg="rtabmap_odom" type="rgbd_odometry" name="rgbd_odometry" output="screen">
<!-- Frames -->
<param name="frame_id" value="base_link"/>
<param name="odom_frame_id" value="odom_vo"/>
<param name="publish_tf" value="false"/>
<!-- Help VO stick using motion guess -->
<param name="Odom/GuessMotion" value="true"/>
<!-- Advanced VO tuning (from Advanced Parameter Tuning) -->
<param name="Odom/Strategy" value="0"/> <!-- F2F odometry -->
<param name="Vis/CorType" value="0"/> <!-- Optical flow -->
<param name="OdomF2M/MaxSize" value="1000"/> <!-- F2M feature buffer -->
<param name="Vis/MaxFeatures" value="1000"/> <!-- per-image features -->
<param name="GFTT/MinDistance" value="40"/> <!-- for ≥720p images -->
<param name="Reg/Force3DoF" value="true"/> <!-- planar odom -->
<!-- Your thresholds/features kept -->
<param name="Odom/MinInliers" value="12"/>
<param name="Vis/MinInliers" value="12"/>
<param name="Kp/MaxFeatures" value="1500"/>
<param name="Odom/FilteringStrategy" value="1"/>
<!-- Sync tuning -->
<param name="approx_sync" value="true"/>
<param name="approx_sync_max_interval" value="0.05"/>
<!-- Auto reset after 1 bad frame -->
<param name="Odom/ResetCountdown" value="20"/>
<!-- Feed EKF odom as guess -->
<remap from="odom" to="/odom"/>
<!-- Camera topics -->
<remap from="rgb/image" to="/camera/rgb/image_color"/>
<remap from="depth/image" to="/camera/depth/image"/>
<remap from="rgb/camera_info" to="/camera/rgb/camera_info"/>
</node>
</group>
<!-- ================== EKF (robot_localization) ================== -->
<group if="$(arg use_ekf)">
<node pkg="robot_localization"
type="ekf_localization_node"
name="ekf_odom"
output="screen">
<remap from="odometry/filtered" to="/odom"/>
<rosparam file="$(find robot_diff)/params/ekf.yaml" command="load"/>
</node>
</group>
<!-- ================== RTAB-Map SLAM ================== -->
<include file="$(find rtabmap_ros)/launch/rtabmap.launch">
<!-- Camera topics -->
<arg name="rgb_topic" value="/camera/rgb/image_rect_color"/>
<arg name="depth_topic" value="/camera/depth_registered/image_raw"/>
<arg name="camera_info_topic" value="/camera/rgb/camera_info"/>
<!-- External odom -->
<arg name="icp_odometry" value="false"/>
<arg name="odom_topic" value="/odom"/>
<!-- Frames -->
<arg name="frame_id" value="base_link"/>
<arg name="odom_frame_id" value="odom"/>
<!-- Sync -->
<arg name="approx_sync" value="true"/>
<arg name="queue_size" value="20"/>
<param name="approx_sync_max_interval" value="0.05"/>
<!-- ===== RTAB-Map (SLAM) params from Advanced Tuning ===== -->
<!-- Planar SLAM (3-DoF) -->
<param name="Reg/Force3DoF" value="true"/>
<param name="Optimizer/Slam2D" value="true"/>
<!-- Initial costmap (min map extent) -->
<param name="grid_size" type="double" value="90"/>
<!-- Reduce 3D point cloud noise (radius outlier removal) -->
<param name="cloud_noise_filtering_radius" value="0.08"/>
<param name="cloud_noise_filtering_min_neighbors" value="100"/>
<!-- Projected occupancy grid characteristics -->
<param name="proj_max_ground_angle" value="60"/>
<param name="proj_max_ground_height" value="0.001"/>
<param name="proj_max_height" value="2.0"/>
<param name="proj_min_cluster_size" value="1000"/>
<!-- Mapping behavior when odom resets -->
<param name="Rtabmap/StartNewMapOnLoopClosure" value="true"/>
<!-- ===== Extra RTAB-Map params for a cleaner, high-fidelity map ===== -->
<!-- Process all frames -->
<param name="Rtabmap/DetectionRate" value="0"/>
<!-- Stronger feature set + stricter matching -->
<param name="Kp/MaxFeatures" value="2500"/>
<param name="Kp/MaxDepth" value="4.5"/>
<param name="Vis/MinDepth" value="0.3"/>
<param name="Vis/MinInliers" value="30"/>
<param name="Vis/CorNNDR" value="0.75"/>
<!-- Refine neighbor constraints + ICP polishing -->
<param name="RGBD/NeighborLinkRefining" value="true"/>
<param name="Icp/PointToPlane" value="true"/>
<param name="Icp/Iterations" value="50"/>
<param name="Icp/MaxCorrespondenceDistance" value="0.05"/>
<!-- Occupancy grid: finer cells + stronger denoising + freespace ray tracing -->
<param name="Grid/CellSize" value="0.02"/>
<param name="Grid/PreVoxelFiltering" value="true"/>
<param name="Grid/NoiseFilteringRadius" value="0.06"/>
<param name="Grid/NoiseFilteringMinNeighbors" value="10"/>
<param name="Grid/NormalsSegmentation" value="true"/>
<param name="Grid/RayTracing" value="true"/>
<!-- Robust loop closures & optimization -->
<param name="Rtabmap/LoopThr" value="0.12"/>
<param name="Optimizer/Robust" value="true"/>
<param name="Optimizer/Iterations" value="200"/>
<param name="RGBD/OptimizeMaxError" value="2.0"/>
<!-- Spatial proximity loop closures -->
<param name="RGBD/ProximityBySpace" value="true"/>
<param name="RGBD/ProximityMaxGraphDepth" value="50"/>
<param name="RGBD/ProximityPathFilteringRadius" value="1.0"/>
<param name="RGBD/ProximityMaxPaths" value="3"/>
<!-- DB + remaining args you already used -->
<arg name="database_path" value="$(arg db_path)"/>
<arg name="rtabmapviz" value="$(arg use_rtabmapviz)"/>
<arg name="rtabmap_args" value="--delete_db_on_start \
--Reg/Strategy 0 \
--Vis/FeatureType 1 \
--Kp/DetectorStrategy 4 \
--Kp/MaxFeatures 1000"/>
</include>
<!-- ================== Virtual Laser (PointCloud2 -> LaserScan) ================== -->
<group if="$(arg virtual_scan)">
<node pkg="pointcloud_to_laserscan" type="pointcloud_to_laserscan_node" name="pc2scan" output="screen">
<param name="target_frame" value="base_link"/>
<param name="min_height" value="0.01"/>
<param name="max_height" value="0.20"/>
<param name="range_min" value="0.01"/>
<param name="range_max" value="8.0"/>
<remap from="cloud_in" to="/camera/depth_registered/points"/>
<remap from="scan" to="/scan"/>
</node>
</group>
</launch>
Which version of rtabmap/ros are you using? rtabmap.launch from rtabmap_ros package (not rtabmap_launch) is kinda old.
Note also that if you launch yourself rgbd_odometry and rtabmap.launch like this, rtabmap.launch also starts its own rgbd_odometry node. You should explicitly set visual_odometry:=false for rtabmap.launch to not use it.
Instead of:
<node pkg="topic_tools" type="relay" name="grid_map_relay" output="screen" args="/rtabmap/grid_map /map" />
you could do:
<remap from="/rtabmap/grid_map" to="/map"/>
before rtabmap.launch include.
To include EKF as guess grame for rgbd_odometry, you cannot do:
<!-- Feed EKF odom as guess -->
<remap from="odom" to="/odom"/>
you should do this instead:
<!-- Feed EKF odom as guess -->
<param name="guess_frame_id" value="odom"/> <!-- the TF frame published by EKF node -->
<remap from="odom" to="/odom_vo"/> <!-- remap to a topic name different than the output /odom from EKF node -->
I don't recommend using:
<param name="Rtabmap/DetectionRate" value="0"/>
the max I set myself is 2 Hz, but generally use always 1 Hz.
For a move_base integration example, you may check this turtlebot3 example on ros1, or this on ros2. Note that in both cases, we don't use visual odometry because indoor with a single camera the visual odometry can get lost often or be not super accurate. If you can get a good wheel / EKF odometry already, I would use only rtabmap node at 1 or 2 Hz update rate. If the occupancy grid looks already good, then to make it smoother and faster it would be more a tuning problem for move_base afterwards. I used rqt_reconfigure a lot with move_base to tune it live, if that can help you.
