Multi-robot Trajectory Planner
This repository contains the code for the paper:
Authors: Juncheng Li, Maopeng Ran, and Lihua Xie from Nanyang Technological University.
Accepted in IEEE Robotics and Automation Letters (RA-L). You can find the full-text paper here.
This paper proposes an efficient trajectory planning approach that generates safe, dynamically feasible and near-optimal trajectories for multiple non-holonomic mobile robots in obstacle-rich environments.
Please click in the image to see our video:
1. Software Requirements
- Ubuntu 16.04 to 20.04
- ROS Kinetic
- Octomap
- Ipopt
2. Installation instructions
(1) Install ROS Kinetic for Ubuntu 16.04
Follow ROS Installation
(2) Install Ipopt solver
Follow Ipopt Installation
(3) Install dependencies
sudo apt-get install ros-kinetic-octomap*
sudo apt-get install ros-kinetic-dynamic-edt-3d
sudo apt-get install cppad
(4) Build:
cd ~/catkin_ws/src
git clone https://github.com/LIJUNCHENG001/multi_robot_traj_planner.git
cd ../ && catkin_make
source ~/catkin_ws/devel/setup.bash
2. Run Simulations
Warehouse
roslaunch multi_robot_traj_planner prioritized_plan_warehouse.launch
Environment with random obstacles
roslaunch multi_robot_traj_planner prioritized_plan_random_env.launch
3. Simulation Configuration
You can configure the simulation settings in the launch files.
(1) Environment: The simulation environment is selected by argument 'replay_map'. The build-in maps are located in /mapfile.
(2) Mission: The mission of the robots is selected by argument 'mission'. The build-in mission files are located in /missions.
(3) Priority Assignment: A prioritized trajectory optimization method is applied to improve the computation efficiency. If argument 'plan_random_group' is true, the priority of the robots is randomly assigned. Otherwise, a novel priority assignment method proposed in the paper is applied.
(4) Backward Movement: If argument 'backward_enable' is true, the robots are able to move backward. Otherwise, the robots can only have forward speed.
4. Acknowledgements
Our implementation is built on top of libMultiRobotPlanning and swarm_simulator. We thank Wolfgang Hönig and Jungwon Park for their great work.