mincheulkim / rlca_trainning

RLCA_trainning

Folder description

• main branch : multiagent
• PPO_R : origin observation ( lidar, goal, speed ) + radius, lidar data ( lidar_(t-2), lidar_(t-1), lidar_(t)
• PPO_R_lidar : PPO_R_new + lidar data ( lidar_(t-10), lidar_(t-5), lidar_(t)
• PPO_R_new : PPO_R + collision lidar add
• PPO_no_update : Do not policy update (Purpose to use as a wander)
• wander : wander code, collision avoidance used scan data
• wander_no_avoidance : wander code
• worlds : stage model file

Setting

Stage & stageros & ppo train py

  sudo apt-get install ros-kinetic-stage
  cd /catkin_ws/src
  git clone https://github.com/Geonhee-LEE/rl-collision-avoidance
  git clone https://github.com/CzJaewan/RLCA_trainning
  cd /catkin_ws
  catkin_make
  cd /catkin_ws/src/rl-collision-avoidance
  git branch stageros_w

Run

stage_run

  roscore
  rosrun stage_ros_add_pose_and_crash stageros_w -u -n 6 /home/nscl/rl_ws/src/RLCA_trainning/worlds/servingbot_single_agent_world

train

  mpiexec -np 1 python ppo_stage.py