Tensor_field_nav package is used for autonomous mapping of unkonown indoor scenes guided by time-varing tensor field. It leverages rgbd sensor like kinect to scan environment, and receives scene geometry information from octomap. Based on the partially mapping scene, a 2D tensor field is computed on the floor plane, under the constraints of 2d grid map. The robot is guided by the field with smooth paths, which are locally formed with advection and globally planned with help of the field topology.
This package relies on some third dependencies, it requires installation of CUDA, Opencv.
Dowload tensor_field_nav package with the following command to your catkin workspace.
catkin_ws/src$ git clone
catkin_ws/src$ cd ..
catkin_ws$ catkin_make
roslaunch tensor_field_nav_core complete_run.launch
roslaunch octomap_tensor_field octomap_mapping.launch