gmapping Mapping Algorithm

1. Course Content

Basics: run the demo program and use the gmapping SLAM algorithm to build a map and save an occupancy grid map.

Advanced: understand the configuration parameters of the gmapping algorithm.

2. gmapping Algorithm Introduction

• gmapping only works when the number of 2D laser points in a single frame is less than 1440. If a single frame contains more than 1440 laser points, a problem like [mapping-4] process has died will occur.
• Gmapping is a commonly used open-source SLAM algorithm based on the filtering SLAM framework.
• Gmapping is based on the RBPF particle filter algorithm, which separates the simultaneous localization and mapping process: localization is performed first, then mapping.
• Gmapping makes two main improvements to the RBPF algorithm: an improved proposal distribution and selective resampling.

Advantages: Gmapping can build indoor maps in real time, and building maps of small scenes requires relatively little computation while achieving high accuracy.

Disadvantages: As the scene grows larger, the number of required particles increases. Because each particle carries a map, both the memory and

computational cost increase when building large maps. Therefore, it is not suitable for mapping large scenes. Moreover, it has no loop-closure detection, so the map may become misaligned when a loop closes; although increasing the number of particles can close the loop, it comes at the cost of higher computation and memory usage.

3. Running the Demo

xxxxxxxxxx
ros2 launch microros_v2_bringup gmapping.launch.py

3.1 Mapping Process

• Note: when mapping, slower speeds give better results (especially the rotation speed should be slow). If you move too fast, the mapping quality will be poor.
• Control the robot to move and complete mapping of the environment. You can use keyboard control or gamepad control; keyboard control is used here as an example.

xxxxxxxxxx
ros2 run common_demo keyboard_control

3.2 Saving the Map

• Run the following command in the terminal to save the occupancy grid map:

xxxxxxxxxx
ros2 launch microros_v2_bringup save_map.launch.py

[!TIP]

Optional parameters:

map_name: specify the name of the map to save; the default value is map

For example, to save a map named test.yaml, add the parameter: ros2 launch microros_v2_bringup save_map.launch.py map_name:=test

A Map saved successfully message in the terminal means the map was saved successfully. If saving fails, try again.

The default map save path is as follows:

xxxxxxxxxx
~/map/map.yaml

One PGM image and one YAML file, yahboom_map.yaml

xxxxxxxxxx
image: yahboom_map.pgm
mode: trinary
resolution: 0.05
origin: [-10, -10, 0]
negate: 0
occupied_thresh: 0.65
free_thresh: 0.25

Parameter explanations:

• image: path to the map file; it can be an absolute path or a relative path
• mode: this attribute can be one of trinary, scale, or raw, depending on the mode chosen; trinary is the default mode
• resolution: the resolution of the map, in meters/pixel
• origin: the 2D pose (x, y, yaw) of the bottom-left corner of the map. Here yaw is rotated counterclockwise (yaw=0 means no rotation). Many parts of the current system ignore the yaw value.
• negate: whether to invert the meaning of white/black and free/occupied (the interpretation of the thresholds is not affected)
• occupied_thresh: pixels with an occupancy probability greater than this threshold are considered fully occupied.
• free_thresh: pixels with an occupancy probability less than this threshold are considered completely free.

3.3 Configuration File

• Configuration file path:

xxxxxxxxxx
~/microros_ws/src/microros_v2_bringup/config/slam_gmapping.yaml

• Meanings of the default configuration parameters:

slam_gmapping:
  ros__parameters:
    # Angular update threshold (radians): the map is updated when the robot rotates beyond this angle
    angularUpdate: 0.25
    # Angular sampling step (radians): angular search step for scan matching
    astep: 0.05
    # Robot base frame name
    base_frame: base_footprint
    # Map frame name
    map_frame: map
    # Odometry frame name
    odom_frame: odom
    # Scan matching iteration termination condition: stop iterating when the parameter change is below this value
    delta: 0.05
    # Maximum number of scan matching iterations
    iterations: 5
    # Kernel size for scan matching, affects matching smoothness
    kernelSize: 1
    
    # Distance sampling range for large-scale scan matching (m)
    lasamplerange: 0.005
    # Distance sampling step for large-scale scan matching (m)
    lasamplestep: 0.005
    # Linear update threshold (m): the map is updated when the robot moves beyond this distance
    linearUpdate: 0.5
    # Distance sampling range for small-scale scan matching (m)
    llsamplerange: 0.01
    # Distance sampling step for small-scale scan matching (m)
    llsamplestep: 0.01
    
    # Measurement noise covariance of the laser scan
    lsigma: 0.075
    # Number of laser beams to skip: 0 means using all laser points
    lskip: 0
    # Linear sampling step (m): linear search step for scan matching
    lstep: 0.05
    # Map update time interval (s)
    map_update_interval: 3.0
    # Maximum laser range (m): points beyond this distance are ignored
    maxRange: 6.0
    # Effective maximum laser range (m): maximum distance used for mapping
    maxUrange: 4.0
    # Minimum score threshold for scan matching: matches below this value are discarded
    minimum_score: 0.0
    # Occupancy probability threshold: cells above this value are considered occupied
    occ_thresh: 0.25
    # Gain factor: affects the map update speed
    ogain: 3.0
    # Number of particles in the particle filter
    particles: 30
    # QoS (Quality of Service) parameter settings
    qos_overrides:
      /parameter_events:
        publisher:
          depth: 1000          # Message queue depth
          durability: volatile # Durability policy
          history: keep_all    # History policy
          reliability: reliable # Reliability policy
      /tf:
        publisher:
          depth: 1000
          durability: volatile
          history: keep_last   # Keep only the latest message
          reliability: reliable
    # Resampling threshold: resampling is triggered when the effective particle count falls below this ratio
    resampleThreshold: 0.5
    # Angular noise covariance of scan matching
    sigma: 0.05
    # Odometry model parameter: linear error ratio caused by linear motion
    srr: 0.1
    # Odometry model parameter: angular error ratio caused by linear motion
    srt: 0.2
    # Odometry model parameter: linear error ratio caused by angular motion
    str: 0.1
    # Odometry model parameter: angular error ratio caused by angular motion
    stt: 0.2
    # Time update interval (s): the map is updated even if the robot does not move
    temporalUpdate: 1.0
    # Coordinate transform publish period (s)
    transform_publish_period: 0.05
    # Whether to use simulated time
    use_sim_time: false
    # Maximum range of the map along the x axis (m)
    xmax: 100.0
    # Minimum range of the map along the x axis (m)
    xmin: -100.0
    # Maximum range of the map along the y axis (m)
    ymax: 100.0
    # Minimum range of the map along the y axis (m)
    ymin: -100.0