Tensor Voting Planner

Prerequsite

• Platform: ROS noetic
• Nvidia Cuda10.1

Run the repo

1. Clone the repo

   git clone https://github.com/GaoLon/TV_planner.git
   

2. Build the workspace

   cd ~/TV_planner
   catkin_make
   

3. Run the demo

   cd ~/TV_planner
   source devel/setup.bash
   roslaunch tensorvoting demo_test_plan.launch
   

   Then you can use /move_base_simple/goal plugin in rviz to set goal and see results of different planner:

   • RED: K-NN
   • GREEN: Shortest path
   • YELLOW: Tensor Voting

Note

1. There are some TODO list:
   • do experiments in different point cloud maps
   • there are some bugs about std::cout or printf function.
   • benchmark
2. Thanks to open-source work done by Liu Ming:
   • Liu, Ming, et al. "Normal estimation for pointcloud using GPU based sparse tensor voting." 2012 IEEE International Conference on Robotics and Biomimetics (ROBIO). IEEE, 2012.
   • Liu, Ming. "Robotic online path planning on point cloud." IEEE transactions on cybernetics 46.5 (2015): 1217-1228.
   • https://ram-lab.com/downloads/