This package contains Gazebo plugins intended to simulate reinforcement learning algorithms in mobile robots.
Supported system: Ubuntu 16.04 Xenial
Basic dependencies: cmake, gcc, opencv
-
Install Gazebo 7
sudo apt-get install libgazebo7-dev
-
Compile and install caffe on the system
cmake .. -DCMAKE_INSTALL_PREFIX=/usr/local
-
Compile and install the source code
mkdir build && cd build cmake .. -DCOMPILE_DRL_PLUGIN=1 make -j4 && sudo make install
- Edit the gazebo world file in gazebo/worlds/
- set the mode tag to test
- select number of iterations in max_steps
- Launch gazebo
gazebo gazebo/worlds/rover.world --verbose gazebo gazebo/worlds/nfq_rover.world --verbose gazebo gazebo/worlds/drl_rover.world --verbose
-
Edit the gazebo world files (rover.world, nfq_rover.world, drl_rover.world):
- set the mode tag to train
- select number of iterations in max_steps
-
Launch gazebo rover world:
gazebo gazebo/worlds/rover.world --verbose
This work is part of Thomio Watanabe PhD project funded by grant: #2015/26293-0, São Paulo Research Foundation (FAPESP). "Opinions, hypothesis and conclusions or recommendations expressed herein are the author(s) responsibility and do not necessarily conform with FAPESP vision."
Copyright © 2017 Thomio Watanabe