PoseHub is a library that manages the transformations of different sensors and objects in the same scene and provides a unified interface for accessing and solving the spatial relationship between frames. For simplification, this package only depends on the numpy and ZeroMQ package. The idea is inspired by the tf library and Open3D.

Clone the repository to the workspace.

git clone https://github.com/jmz3/PoseHub.git

If you want to use the ros features of the package, you need to build the ros-related packages, namely posehub_ros and posehub_tools.

cd PoseHub
catkin build

But if you only want to use the core features of the package, no more installation is needed. You can directly use the package by importing the PoseGraph class.

The package is organized as follows:

src
├── PoseHub       # ros-free core implementation of the posehub
├── posehub_ros   # ros wrapper for the posehub
└── posehub_tools # synthetic data generator for testing

Open a terminal and run the following command to start the ZMQ server. Note that you need to modify the IP address in the posehub_main.py file to the IP address of your sensor (either HoloLens or something else).

python3 posehub_main.py

If you have multiple sensors, you can copy the following code snippet and modify the args to create multiple communication objects.

args_1 = argparse.Namespace(
    sub_ip=args.sub_ip_1,
    sub_port="5588",
    pub_port="5589",
    sub_topic=[tool_1_id, tool_2_id, tool_3_id],
    pub_topic=[tool_1_id, tool_2_id, tool_3_id],
    sensor_name=sensor_1_id,
)  # args for h1 sensor

# create zmq manager
zmq_manager_1 = ZMQManager(
    sub_ip=args_1.sub_ip,
    sub_port=args_1.sub_port,
    pub_port=args_1.pub_port,
    sub_topic=args_1.sub_topic,
    pub_topic=args_1.pub_topic,
    sensor_name=args_1.sensor_name,
)

# initialize zmq manager
zmq_manager_1.initialize() 

# create pose graph object to store the transformation
pose_graph = PoseGraph()
pose_graph.add_sensor("h1")

Since our communication objects are running in different threads, we need to join the threads by calling the terminate function when we want to exit the program.

# join the threads
zmq_manager_1.terminate()

The graph can be visualized by calling the vis_graph function in the realtime (max 100Hz). Here is an example of the visualization of the graph. The red and green markers represent the frame is being tracked by different sensors.

The synthetic data generator can create "real" frame motion and noise. To use the synthetic data generator, you can run the following command in the terminal.

roslaunch posehub_tools synthetic.launch

For pure trajectory data, you can run the following command.

roslaunch posehub_tools fake_tf.launch

Here is how the synthetic data looks like in the rviz.

To visualize and save the synthetic trajectory data, use the ros node posehub_tools/trajectory_visualizer.py. The following command will save the trajectory data to the posehub_ros/outputs folder.

rosrun posehub_ros VizTrajectory.py

• Modularized TCP/IP communication for receiving data from different sensors
• Unified Graph structure for storing the transformation between different sensors and objects
• Graph search based spatial transformation solver
• Random synthetic data generator, and noise injection for testing

The communication objects are running in different threads. They are created and started in the main thread. The main thread is responsible for creating and starting the communication objects. The communication objects are responsible for receiving data from the sensors and updating the PoseGraph object. The PoseGraph object is declared in the main thread so that it can be accessed by all the communication objects. The PoseGraph object is locked when it is being updated by the communication objects. The PoseGraph object is unlocked when it is being accessed by the main thread.

The package is designed to be used in the Surgical Tool Tracking using Multi-Sensor System, the final project of EN.601.654: Augmented Reality. The contributors of this package are:

• Jiaming Zhang
• Hongchao Shu