This is a Ros Version Mad-ICP

1. ROS support,with a simple viewer
2. multi-type lidar support(velodyne,ouster,robosense,and pandar)

A minimal, robust, accurate, and real-time LiDAR odometry.

Building has been tested on Ubuntu 20.04 (with g++).

The following external dependencies are required.

After obtaining all dependencies, the application can be built with CMake, for example, as follows:

cd mad-icp
mkdir build && cd build && cmake .. && make -j

A few other Python packages need to be installed for running, namely: numpy, pyyaml, and rosbags. You can find the specific versions in the requirements.txt. We suggest to create a virtual env and run pip3 install -r requirements.txt.

We provide a Python launcher for both rosbags and bin formats (we are currently working on bin format luncher; for this reason, you do not find configurations for KITTI and Mulran). The configuration file is important for the sensor characteristics and extrinsic information (most of the time, ground truths are not in the LiDAR frame). The internal parameters are in configurations/params.cfg; all the experiments have been run with this same set.

How to run (rosbag), make sure estimate_path and data_path point to a folder:

cd apps
python3 rosbag_runner.py --data_path /path_to_bag_folder/ --estimate_path /path_to_estimate_folder/ --dataset_config ../configurations/datasets/dataset_config_file --mad_icp_config ../configurations/params.cfg --num_cores 4 --num_keyframes 4 --realtime

Our runner directly saves the odometry estimate file in KITTI format (homogenous matrix row-major 12 components); in the near future, we will provide more available formats like TUM. Our pipeline is anytime realtime, therefore you can play with parameters num_keyframes and num_cores, if you have enough computation we suggest to increase these (we run demo with 16 and 16), if not you can leave it in the proposed way.