ndt_2d

This package implements the Normal Distribution Transform (NDT) for mapping and localization.

The scan matcher is plugin-based. An example of using the Karto scan matcher can be found in the ndt_2d_karto package. This opens up some interesting applications, such as mapping with the Karto scan matcher and then using the NDT-based one for localization afterwards.

This package is still a work-in-progress - see the open issue for the best estimate of what does or does not work.

Use Cases

• Mapping: default parameters will lead to this use case. The save_map.py script can be used to save the NDT map data for later localization. The regular nav2_map_server can be used to save the map needed for navigation, or it can be re-generated at runtime from the NDT map data.

• Localization via Particle Filter: set the map_file parameter to the full filename of your saved NDT map data, set use_particle_filter to true. The particle filter supports many of the same odometry and filter parameters as AMCL (e.g. odom_alpha1, min_particles);

• Localization via Scan Matching: set the map_file parameter to the full filename of your saved NDT map data, make sure use_particle_filter is set to false.

Parameter Details

• enable_mapping: When set, mapping is disabled. A global NDT will be built from the loaded map.

• global_search_size: The maximum distance between two scans to be considered for global loop closure.

• global_search_limit: The maximum number of scans to be considered for global loop closure against a new scan.

• minimum_travel_distance: Minimum linear travel distance before localization update is applied. Applies to both particle filter and scan matching based localization. Units: meters.

• minimum_travel_rotation: Minimum angular travel before localization update is applied. Applies to both particle filter and scan matching based localization. Units: radians.

• map_file: If this set, this resource will be loaded as an initial map. This works for both continuing to map OR localization. Robot must be localized with the initial pose tool.

• max_range: Maximum distance of laser measurements. Measurements beyond this range are discarded. Default is -1, in which case the max range will be extracted from the laser scan message.

• occupancy_threshold: When generating the occupancy grid map, this is the threshold between free and occupied space based on how many raytraces have hit or passed through a given cell.

• odom_frame: TF frame_id for the odometry. Usually odom.

• optimization_node_limit: Minimum number of nodes that must be added to the graph between runs of the graph optimizer.

• resolution: Resolution of the published occupancy grid map. This is entirely independent of the underlying resolution of the NDT. Units: meters.

• robot_frame: TF frame_id for the robot. Usually base_link.

• rolling_depth: When building a map, this is how many scans to use when building the local NDT for scan matching.

• scan_matcher_type: The plugin name for the scan matcher to use. Default is ndt_2d::ScanMatcherNDT.

• transform_timeout: Max allowable time to wait for transform to become available when transforming the laser scan. Units: seconds.

• use_barycenter: When scan matching, should closest scans be selected via the scan pose or the barycenter of the scan points.

• use_particle_filter: When set, mapping is disabled and a global NDT is created. The initial pose tool will initialize localization.

ScanMatcherNDT Parameters

Each scan matcher uses the following parameters, namespaced into either local_scan_matcher or global_scan_matcher namespaces:

• ndt_resolution: Resolution used for the NDT grid. Every cell of this resolution will be represented by a single Gaussian function. Units: meters.

• laser_max_beams: Maximum number of laser beams to use during scan matching. This mirrors the parameter of the same name in AMCL.

• search_angular_resolution: Angular resolution to use for the scan matching search. Units: radians.

File truncated at 100 lines see the full file