HATSDF SLAM

HATSDF SLAM
- Dependencies
- Build
- Base Design
- Runtime Parameters
- Run
- Cite

Dependencies

Ubuntu 18.04
make
Xilinx 2020.1
Hardware Base Design Pre-Built | DIY
qemu (Tests)

Build

Note: HW_TARGET=sw_emu is now default, compile hardware with HW_TARGET=hw

Source Xilinx Tools: source <XILINX_INSTALL_DIR>/settings64.sh
Source Xilinx Runtime: source /opt/xilinx/xrt/setup.sh
Are specific board files copied? In FastSense repo, run cp -r base_design/board_files/* <VIVADO_INSTALL_DIR>/data/boards/board_files/
Compile project
- software only? make software -j4
- hardware only? make hardware -j4
- tests (hw+sw)? make test -j4
- test software only? make test_software
- test hardware only? make test_hardware
Package SD card image (if HW_TARGET=sw_emu qemu launch script and image is generated)
- HATSDF SLAM: make package HW_TARGET=hw -j4
- Tests: make package_test HW_TARGET=hw -j4

Base Design

You can either download a pre built base-design for the specific board, or build it yourself. To do just that please follow the instructions in the Base Design repository.

Runtime Parameters

All parameters are read when the program is started from the config.json located in app_data/. After copying the files to the SD card, this file should be copied into the /mnt directory alongside the executable.

In the following, the runtime parameters are listed and described shorly:

imu: Parameters for the IMU driver and the IMU average filter
- bufferSize: Buffer size for the incoming imu data
- filterSize: Size of the sliding window for the average filter
lidar: Parameters for the LiDAR driver to configurate the network communication and the algorithm based on the properties of the laserscanner
- bufferSize: Buffer size for the incoming LiDAR data
- port: Port for the network communication of the sensor
- pointscale: Scaling that should be applied on every point of the received cloud. This can indirectly adjust the resolution of the map without changing the hardware
- rings: Expected scan rings in the received point clouds
- vertical_fov_angle: Expected vertical field of view of the sensor (in degree)
registartion: Parameters for the registration of the incoming sensor data based on the current TSDF map
- max_iterations: Maximum number of iterations spent for every matching procedure
- it_weight_gradient: Weight for the changing decay, increasing with every iteration
- epsilon: Registartion error from which the matching procedure should be stopped
gpio: Parameters for the GPIO pins
bridge: Parameters for the ROS bridge
- use_from: Should the the sensor data be used from the ROS bridge?
- use_to: Should the output data of the SLAM-Box be sent via the ROS bridge?
- send_original: Send the unprocessed sensor data provided by the laserscanner
- send_preprocessed: Send the scaled and filtered sensor data
- send_after_registartion*: Send the registered sensor data
- host_from: IP address of the host
- imu_port_from: Port to receive IMU data from the host
- imu_port_to: Port to send IMU data send data to the host
- pcl_port_from: Port to receive point cloud data from the host
- pcl_port_to: Port to send point cloud data to the host
- transform_port_to: Port to send the current pose of the SLAM-Box to the host
- tsdf_port_to: Port to send the local TSDF map to the host
slam: Parameters for the mapping procedure
- max_distance: Truncation value for the distance values in the TSDF map (in mm)
- map_size_x: Size of the local TSDF map in x direction (in cells)
- map_size_y: Size of the local TSDF map in y direction (in cells)
- map_size_z: Size of the local TSDF map in z direction (in cells)
- max_weight: Upper bound for the weights of every cell for the averaging
- initial_map_weight: Initial weight for every cell in the TSDF map
- map_update_period: Skipped scans until the next map update
- map_update_position_threshold: Distance from which a new map update is to be performed
- map_path: Save directory for the global map

Example:

{
    "imu": {
        "bufferSize": 16,
        "filterSize": 25
    },

    "lidar": {
        "bufferSize": 1,
        "port": 2368,
        "pointScale": 1.0,
        "rings": 16,
        "vertical_fov_angle": 30.0
    },

    "registration": {
        "max_iterations": 200,
        "it_weight_gradient": 0.1,
        "epsilon": 0.01
    },

    "gpio": {
        "button_chip": "gpiochip0",
        "button_line": 0,
        "led_chip": "gpiochip1",
        "led_line": 0
    },

    "bridge": {
        "use_from": true,
        "use_to": true,

        "send_original": false,
        "send_preprocessed": false,
        "send_after_registration": true,

        "host_from": "192.168.1.245",

        "imu_port_from": 4444,
        "imu_port_to": 5555,

        "pcl_port_from": 3333,
        "pcl_port_to": 7777,

        "transform_port_to": 8888,
        "tsdf_port_to": 6666
    },

    "slam": {
        "max_distance": 600,
        "map_size_x": 201,
        "map_size_y": 201,
        "map_size_z": 121,
        "max_weight": 10,
        "initial_map_weight": 0.0,
        "map_update_period": 100,
        "map_update_position_threshold": 500,
        "map_path": "/data"
    }
}

Run

On the SoC:

cd /mnt
./FastSense.exe

Cite

@inproceedings{eisoldt2021hatsdf,
  title={HATSDF SLAM--Hardware-accelerated TSDF SLAM for Reconfigurable SoCs},
  author={Eisoldt, Marc and Flottmann, Marcel and Gaal, Julian and Buscherm{\"o}hle, Pascal and Hinderink, Steffen and Hillmann, Malte and Nitschmann, Adrian and Hoffmann, Patrick and Wiemann, Thomas and Porrmann, Mario},
  booktitle={2021 European Conference on Mobile Robots (ECMR)},
  pages={1--7},
  year={2021},
  organization={IEEE}
}

@inproceedings{flottmann2021energy,
  title={Energy-efficient FPGA-accelerated LiDAR-based SLAM for embedded robotics},
  author={Flottmann, Marcel and Eisoldt, Marc and Gaal, Julian and Rothmann, Marc and Tassemeier, Marco and Wiemann, Thomas and Porrmann, Mario},
  booktitle={2021 International Conference on Field-Programmable Technology (ICFPT)},
  pages={1--6},
  year={2021},
  organization={IEEE}
}

@article{eisoldt2022fully,
  title={A fully integrated system for hardware-accelerated TSDF SLAM with LiDAR sensors (HATSDF SLAM)},
  author={Eisoldt, Marc and Gaal, Julian and Wiemann, Thomas and Flottmann, Marcel and Rothmann, Marc and Tassemeier, Marco and Porrmann, Mario},
  journal={Robotics and Autonomous Systems},
  volume={156},
  pages={104205},
  year={2022},
  publisher={Elsevier}
}

@incollection{eisoldt2021reconfros,
  title={Reconfros: Running ros on reconfigurable socs},
  author={Eisoldt, Marc and Hinderink, Steffen and Tassemeier, Marco and Flottmann, Marcel and Vana, Juri and Wiemann, Thomas and Gaal, Julian and Rothmann, Marc and Porrmann, Mario},
  booktitle={Proceedings of the 2021 Drone Systems Engineering and Rapid Simulation and Performance Evaluation: Methods and Tools Proceedings},
  pages={16--21},
  year={2021}
}

@article{wiemannenergieeffizientes,
  title={Energieeffizientes FPGA-basiertes TSDF-SLAM f{\"u}r Mobile Mapping},
  author={WIEMANN, Thomas and EISOLDT, Marc and FLOTTMANN, Marcel and GAAL, Julian and ROTHMANN, Marc and TASSEMEIER, Marco and PORRMANN, Mario}
}