AI-BS Summer School 2017

AI-BS Summer School 2017 is to provide opportunities for young researchers to meet each other and learn recent studies in the field of Artificial Intelligence and Brain Science. In particular, this project gives you to enjoy deep reinforcement learning for robot control.

ROS and OpenAI Gym

~$ git clone https://github.com/uchibe/ai-bs-summer17.git
$ cd ai-bs-summer17
ai-bs-summer17$ ./run ros_cpu

Now, you are in the container.

Roboschool

Pretrained model

python3 $ROBOSCHOOL_PATH/agent_zoo/RoboschoolHumanoidFlagrun_v1_2017jul.py

python3 $ROBOSCHOOL_PATH/agent_zoo/demo_race2.py

Baselines

OpenAI Baselines is a set of implementations of reinforcement algorithms. In the container, use python3 to run the sample prorams as follows:

$ python3 -m baselines.deepq.experiments.train_cartpole

Load the model saved in cartpole_model.pkl and visualize the learned policy

python3 -m baselines.deepq.experiments.enjoy_cartpole

python3 -m baselines.deepq.experiments.train_pong

python3 -m baselines.deepq.experiments.enjoy_pong