This repository provides height_mapping ROS package for 2.5D gridded height mapping (or elevation mapping), designed for mobile robots operating The height_mapping package creates a dense terrain map around the robot, which is necessary for the downstream tasks like traversability estimation, obstacle avoidance, and path planning in challenging rough terrain.

This is a research code, expect that it changes often and any fitness for a particular purpose is disclaimed.

Overview: The package is tailored for ground mobile robots, equipped with range sensors (e.g. structured light (Kinect, RealSense), LiDAR, stereo camera) and a 3D pose estimator (e.g. with 3D Visual / LiDAR Odometry or SLAM system). Given 3D pose estimation and range measurements, the package creates a dense terrain map around the robot. During the navigation of a robot, the precise geometry of surrounding terrains are estimated and mapped in a robot-centric perspective. This is crucial for the tasks like traversability estimation, obstacle avoidance, and path planning in challenging rough terrain.

Here are some several key features that the height_mapping package provides:

• Efficient 2.5D Grid Map Structure: Utilizing a 2.5D grid map structure, the package efficiently manages and processes 3D terrain geometry, balancing detail with computational load. This structure is particularly well-suited for real-time applications, enabling quick updates and adjustments as the robot navigates.

• Multiple Height Estimation Methods: The height_mapping node offers several height estimation techniques, including Statistical Mean Filter, Kalman Filter, and Moving Average Filter. This flexibility allows the package to adapt to various mapping needs and environmental conditions.

• Multi-Sensor Fusion: The sensor_processor node integrates data from multiple range sensors (e.g., LiDAR, structured light sensors like Kinect and RealSense, and stereo cameras). This fusion results in a comprehensive and accurate representation of the surrounding terrain, enhancing much denser representation of the environment with larger field-of-view.

• Global Mapping: The global_mapping node aggregates local height maps into an extended global map, providing a broader context for navigation. This global perspective is crucial for long-range planning and overall mission success.

Dependencies: This software is built on the Robotic Operating System (ROS). We assume that the followings are installed.

• Ubuntu (Tested on 20.04)
• ROS (Tested on ROS Noetic)
• PCL (Point cloud library, tested on pcl-1.10)
• grid_map library

For the installation of grid_map, use the following commands:

sudo apt install ros-noetic-grid-map
sudo apt install ros-noetic-grid-map-visualization

Build: In order to install height_mapping package, clone the latest version from this repository and compile the package from source.

cd ~/{your-ros-workspace}/src
git clone https://github.com/Ikhyeon-Cho/height_mapping.git
cd ../
catkin build height_mapping

Note: For the best performance, complie with the option -DCMAKE_BUILD_TYPE=release. It makes significant improvements.

1. Configure the parameters in height_mapping_ros/config/params.yaml
2. To start the height mapping, use command

roslaunch height_mapping run.launch

The sensor_processor node do some sensor input pre-processing for height mapping. The pre-procssing step includes:

• Downsampling of point clouds
• Range-based Filtering of point clouds
• Transformation of point clouds to the robot frame
• Merging multiple sensor inputs

• /sensor1/points (sensor_msgs/PointCloud2)
  Raw range measurements from sensor 1 (Both LiDAR and RGB-D are supported). To configure the name and the type of the measurements, set the corresponding parameters in config/params.yaml.

• /sensor2/points (sensor_msgs/PointCloud2)
  Raw range measurements from sensor 2 (Both LiDAR and RGB-D are supported).

• /sensor3/points (sensor_msgs/PointCloud2)
  Raw range measurements from sensor 3 (Both LiDAR and RGB-D are supported).

• /tf (tf2_msgs/TFMessage)
  Transforms from tf tree. The sensor inputs will be transformed to the robot frame (typically, base_link).

• /height_mapping/sensor/laser/points (sensor_msgs/PointCloud2)
  If LiDAR sensor is used as an input, then the preprocessed pointcloud is published with this name. The messages are fed into the height_mapping node for terrain mapping.

• /height_mapping/sensor/color/points (sensor_msgs/PointCloud2)
  If RGB-D sensor is used as an input, then the preprocessed pointcloud is published with this name. The messages are fed into the height_mapping node for terrain mapping.

• sensor_processor/pointType (string, default: "laser" )
  The type of range sensors to be used. Choose from ["laser", "color"].

• sensor_processor/inputCloudTopic1 (string, default: "/sensor1/points")
  You can configure the topic name of the input measurement streams here. Currently, maximum three sensor inputs are supported.

• sensor_processor/inputCloudTopic2 (string, default: "/sensor2/points")
  You can configure the topic name of the input measurement streams here.

• sensor_processor/inputCloudTopic3 (string, default: "/sensor3/points")
  You can configure the topic name of the input measurement streams here.

• sensor_processor/downsamplingResolution (double, default: 0.0)
  The downsampling resolution of the pointcloud. If set to 0.0, then the processor skips the downsampling. Unit: [m/voxel]

• sensor_processor/minRangeThreshold (double, default: 0.3)
  The minimum range of the measurements from the robot. The points less than the range will be removed and not used for height mapping. Unit: [m]

• sensor_processor/maxRangeThreshold (double, default: 10.0)
  The maximum range of the measurements from the robot. The points over the range will be removed and not used for height mapping. Unit: [m]

• sensor_processor/cloudPublishRate (double, default: 15.0)
  The publish rate of the processed pointclouds that will be fed into the mapping process.

The height_mapping node that locally maps the surrounding terrains in a robot-centric perspective.

• /height_mapping/sensor/laser/points (sensor_msgs/PointCloud2)
  The pre-processed LiDAR sensor input from sensor_processor node.

• /height_mapping/sensor/color/points (sensor_msgs/PointCloud2)
  The pre-processed RGB-D sensor input from sensor_processor node.

• /height_mapping/map/gridmap (grid_map_msgs/GridMap)
  The estimated height map.

• frame_id/map (string, default: "map" )
  The frame id of the height map is coincide with the map frame.

• height_mapping/heightEstimatorType (string, default: "StatMean" )
  The height estimation methods. Choose from ["StatMean", "KalmanFilter", "MovingAverage"].

• height_mapping/gridResolution (double, default: 0.1 )
  The grid resolution of the height map. Unit: [m/grid]

• height_mapping/mapLengthX (double, default: 10.0 )
  The Length of height map in the x-axis of the map frame. Unit: [m]

• height_mapping/mapLengthY (double, default: 10.0 )
  The Length of height map in the y-axis of the map frame. Unit: [m]

• height_mapping/minHeightThreshold (double, default: -0.5 )
  This parameter defines the minimum height below the ground surface that will be considered in the map. Any height below this threshold will be disregarded. Unit: [m]

• height_mapping/maxHeightThreshold (double, default: 1.5 )
  This parameter defines the maximum height above the ground surface that will be considered in the map. Any height above this threshold (ex. overhanging obstacles) will be disregarded. Unit: [m]

• height_mapping/poseUpdateRate (double, default: 15.0 )
  The pose update rate. Height map will follow the robot poses as it is updated. The pose is updated from tf tree. Unit: [Hz]

• height_mapping/mapPublishRate (double, default: 15.0 )
  The height map publish rate. Unit: [Hz]

The map_visualization node that publishes the several height map-related messages, like map regions or height map in pointclouds formats.

• /height_mapping/map/gridmap (grid_map_msgs/GridMap)
  The estimated height map.

• /height_mapping/map/pointcloud (sensor_msgs/PointCloud2)
  The estimated height map in point cloud types.

• /height_mapping/map/region (visualization_msgs/Marker)
  The boundary of the height map.

The global_mapping node subscribes the local height map (in pointcloud) from the height_mapping node, and publishes the map in a much large (global) scale.

• /height_mapping/map/pointcloud (sensor_msgs/PointCloud2)
  The estimated height map in point cloud types.

• /height_mapping/globalmap/pointcloud (sensor_msgs/PointCloud2)
  The estimated global height map in point cloud types.

• /height_mapping/globalmap/region (visualization_msgs/Marker)
  The boundary of the global height map.

• global_mapping/gridResolution (double, default: 0.1 )
  The grid resolution of the global height map. Unit: [m/grid]

• global_mapping/mapLengthX (double, default: 400.0 )
  The Length of global height map in the x-axis of the map frame. Unit: [m]

• global_mapping/mapLengthY (double, default: 400.0 )
  The Length of global height map in the y-axis of the map frame. Unit: [m]

• global_mapping/mapPublishRate (double, default: 10.0 )
  The global height map publish rate. Unit: [Hz]

• global_mapping/mapSaveDir (string, default: "{HOME}/Downloads" )
  The folder path to save global map in image.

To see the real-world application of this project, please refer to:

International:

• 'Learning Self-supervised Traversability with Navigation Experiences of Mobile Robots: A Risk-aware Self-training Approach' (IEEE Robotics and Automation Letters, paper accepted in Feb., 2024).

Domestic:

• 'Traversability Analysis in Outdoor Environment with Elevation Grid Map Update Method' (Conference on Information and Control Systems, in Oct., 2021)