ndt_mapping

ndt_mapping copied to clipboard

Quick Start

This is Baidu apollo offline mapping tool.

Environment

The environment for creating the map is as follows, you need to be equipped with lidar and GNSS(IMU+GPS).

• RoboSense RS-LiDAR-32
• GNSS
• Apollo 5.0

Pls use Apollo 5.0 for now !!! 6.0 causes a segment fault

Created branches with different apollo versions, please pull the branch based on the apollo version. eg: apollo 5.0 pull r5.0.0, apollo6.0 pull r6.0.0.

Collect data

First you need to collect the sensor data needed for mapping. If your vehicle has been installed with Apollo5.0, you can use below command to record the bag.

cyber_recorder record -c imu_topic localization_pose_topic lidar_topic

After collecting the data, you can start making a map by following the steps below.

How to run

Compile the code according to the following steps.

1. Copy the entire directory ndt_mapping to apollo's modules/localization/msf/local_tool/
2. Build the localization module in apollo

./apollo.sh build localization   // apollo 6.0

// or compile separately

bazel build modules/localization/msf/local_tool/ndt_mapping

It should be noted that using "bazel build xxx" will clear the compilation cache, and the next time you need to compile apollo because there is no cache, it will be recompiled, which is very time-consuming.

1.Unzip the bag

Extract the pcd file and pose file from the bag. You can use multiple "--bag_file" to extract multiple bag files. The decompressed file is saved in --out_folder.

./bazel-bin/modules/localization/msf/local_tool/data_extraction/cyber_record_parser --bag_file=data/bag/20210514100819.record.00000 --bag_file=data/bag/20210514100819.record.00001 --out_folder=data --cloud_topic=/apollo/sensor/rs32/PointCloud2

2.Poses interpolation

Interpolate the pose according to the external parameters and timestamp of the lidar. The corrected pose is saved in --output_poses_path.

./bazel-bin/modules/localization/msf/local_tool/map_creation/poses_interpolator --input_poses_path=data/pcd/odometry_loc.txt --ref_timestamps_path=data/pcd/pcd_timestamp.txt --extrinsic_path=modules/localization/msf/params/velodyne_params/velodyne64_novatel_extrinsics_example.yaml --output_poses_path=data/pcd/poses.txt

3.NDT mapping

Use the following command to create a map, the result of the map is default saved in "data/output.pcd"

./bazel-bin/modules/localization/msf/local_tool/ndt_mapping/ndt_mapping

You can also set some parameters in the command line according to your needs

./bazel-bin/modules/localization/msf/local_tool/ndt_mapping/ndt_mapping -output_file=xxx

The parameters list

// filter
-min_scan_range = 25.0  // the square of the min scan range
-max_scan_range = 10000.0  // the square of the max scan range
-min_add_scan_shift = 1.0  // the square of the min add scan length
-voxel_leaf_size = 2.0  // voxel leaf size

// ndt
-trans_eps = 0.01  // transformation epsilon
-step_size = 0.1  // step size
-ndt_res = 1.0    // ndt resolution
-max_iter = 30    // maximum iterations times

// map
-output_file = "data/output.pcd"  // map save file path
-workspace_dir = "data/pcd"       // work dir

Example

The following is the result of the mapping of the underground parking lot.