⚠️ This repository is archived. For the latest version of RLSS, please email baskin.senbaslar@usc.edu.

RLSS: Distributed Real-time Trajectory Replanning using Linear Spatial Separations

RLSS is a real time trajectory replanning algorithm for multi robot teams. RLSS combines A*-based discrete search with QP based trajectory optimization where linear separations enforce safety.

RLSS,

• computes provably executable trajectories for differentially flat robots
• enforces safety as hard constraints
• can work in real-time (1-10 Hz)
• requires perfect sensing of only the positions of robots and obstacles
• does not depend on communication for safety

Citing RLSS

Please use the following entry to cite RLSS:

@article{senbaslar2021rlss,
    author = {{{\c{S}}enba{\c{s}}lar}, Bask{\i}n and {H{\"o}nig}, Wolfgang and {Ayanian}, Nora},
    title = "{RLSS: Real-time Multi-Robot Trajectory Replanning using Linear Spatial Separations}",
    journal = {arXiv e-prints},
    keywords = {Computer Science - Robotics},
    year = 2021,
    month = mar,
    eid = {arXiv:2103.07588},
    pages = {arXiv:2103.07588},
    archivePrefix = {arXiv},
    eprint = {2103.07588},
    primaryClass = {cs.RO}
}

Dependencies

RLSS depends on Eigen3, Boost. It also depends on ILOG CPLEX C++, GUROBI as optimization libraries.

Building

git clone -b master https://github.com/usc-actlab/rlss/
git submodule update --init --recursive
mkdir build
cd build
cmake -DCMAKE_BUILD_TYPE=Release -DBUILD_RLSS_EXAMPLES=ON -DENABLE_RLSS_JSON_BUILDER=ON ..
make -j4

If you are having problems linking Gurobi

cd $GUROBI_HOME/src/build
make
cd $GUROBI_HOME/lib
rm libgurobi_c++.a
ln -s ../src/build/libgurobi_c++.a  libgurobi_c++.a

Targets

• make 2d_sim: build 2D simulator example
• make 3d_sim: build 3D simulator example
• make build_rlss_examples: 2d_sim and 3d_sim
• make build_rlss_tests: build tests
• make test: run tests

Running Simulator Examples & Visualization

• Set base_path to the rlss root folder in examples/2d_config.json and examples/3d_config.json
• Dependencies for 2D visualizer

  pip3 install matplotlib
  sudo apt install ffmpeg
  pip3 install numpy
  

• For 2D visualization run:

  cd build
  ./2d_sim (outputs vis.json)
  python3 ../tools/vis/2d/run.py vis.json
  

• 3D visualizer requires ROS and rviz.
• Dependencies for 3D visualizer

  pip3 install rospkg
  pip3 install sympy
  pip3 install numpy
  BEZIER_INSTALL_PREFIX=./ BEZIER_NO_EXTENSION=true pip3 install bezier==2020.5.19
  

• For 3D visualizer run:

  (terminal A) cd build
  (terminal A) ./3d_sim (outputs vis.json)
  (terminal B) roscore
  (terminal C) rosrun rviz rviz
  (rviz) Add > MarkerArray (topic: visualization_marker_array)
  (rviz) Set background color to white
  (terminal A) python3 ../tools/vis/3d/run.py
  (terminal A) start vis.json (reads vis.json and displays the result in rviz)
  

3D Visualizer Commands

3D visualizer accepts several command line commands for visualization.

• start <simulation_output>: Starts the visualization for the simulation output file
• stop: Stops the current visualization
• forward: Run visualization forward
• backward: Run visualization backward
• pause: Pause visualization
• step: Move 1 step in visualization
• unpause: Unpause visualization
• help: List available commands
• exit: Exit visualizer

ROS Integration

The ROS integration of RLSS is developed in a separate repo: rlss_ros.