Line trace robot simulation with reinforcement learning, Deep Q-Network.
The final report of "Intelligent Informatics", Creative Informatics, the University of Tokyo.
- Ubuntu 18.04
- ROS Melodic
- Python 3.6.x (must be installed on the system like
/usr/bin/python3)
Clone packages.
$ mkdir -p ~/catkin_ws/src
$ cd ~/catkin_ws/src
$ wstool init .
$ wstool set --git line_trace_dqn [email protected]:ketaro-m/line_trace_dqn.git -y
$ wstool update
$ wstool merge -t . line_trace_dqn/.rosinstall
$ wstool update Build packages.
$ cd ..
$ rosdep install -y -r --from-paths src --ignore-src
$ source /opt/ros/${ROS_DISTRO}/setup.bash
$ catkin build line_trace_dqn
$ source ./devel/setup.bash$ roslaunch line_trace_dqn turtlebot3_linetrace.launch $ roslaunch line_trace_dqn test_rospy.launch script:=follower $ roslaunch line_trace_dqn result_graph.launch # realtime score/action plot
$ # roslaunch line_trace_dqn turtlebot3_dqn_train.launch --ros-args # see the parameter descriptions
$ roslaunch line_trace_dqn turtlebot3_dqn_train.launch lr:=0.1 $ rosrun line_trace_dqn log_plotter.py