deep_motion_planning

Using Deep Neural Networks for robot navigation purposes:

The repository contains various ROS packages that are used to generate the training data.

Usage

With Move Base

In order to start the simulation, the navigation and load the static map, run:

roslaunch stage_worlds shapes.launch deep_motion_planning:=False

You can then start a mission by executing:

rosrun mission_control mission_control_node.py _mission_file:=src/deep_motion_planning/mission_control/missions/rooms.txt _deep_motion_planner:=False

The parameter <_mission_file> defines the path to the mission definition that you want to execute. The example above assumes that you are in the top folder of your catkin workspace.

Finally, if you want to see a visualization of the system, run:

roslaunch stage_worlds rviz.launch

With Deep Motion Planner

To start the simulation in combination with the deep motion planner, run:

roslaunch stage_worlds shapes.launch deep_motion_planning:=True

You can then start a mission by executing:

rosrun mission_control mission_control_node.py _mission_file:=src/deep_motion_planning/mission_control/missions/rooms.txt _deep_motion_planner:=True

The parameter <_mission_file> defines the path to the mission definition that you want to execute. The example above assumes that you are in the top folder of your catkin workspace.

Packages

stage_worlds

The stage_worlds package contains the configuration files for the simulation, various world definitions and their ROS launch files.

mission_control

This package contains a mission control node which executes a user defined mission. Therefor, a txt file is parsed and a sequence of commands is generated. This sequence is then processed step-by-step. For more details on the definition of a mission, please refer to the README file in the package directory.

data_capture

The data_capture node subscribes to a set of topics and writes the time-synchronized messages into a .csv file. Currently, the node records data from a laser scanner, the relative target pose for the navigation and the control commands that are send to the robot.

deep_motion_planner

The package wraps a Tensorflow model for motion planning with a deep neural network. The node loads a pretrained model and computes the control commands for the robot from raw sensor data.

deep_learning_model

This is not a ROS package, but a independent project to train a Tensorflow model on data that is captured with the data_capture package. The resulting trained model can then be loaded and executed in the deep_motion_planner package