Code implementation for tightly and effiiciently coupling imitation learning and reinforcement learning to derive mobile robot goal-driven navigation under minimum demonstrated data, meanwhile achieving better sample efficiency and training safety.
Required software ubuntu 14.04 ros-indigo gazebo 7.0
sudo cp -r ~/catkin_ws ./ cd catkin_ws && catkin_make
sudo apt-get install ros-kinetic-*
cd ~/UGV_navigation_RL/record_action && python record_action.py
cd ~/UGV_navigation_RL/offline_training && python offline_training.py
cd ~/UGV_navigation_RL/train_human_exp && python train_human_exp1.py
cd ~/UGV_navigation_RL/real_test && python train_human_exp.py