Learning to Switch Between Machines and Humans

Requirements

To install requirements:

pip install -r requirements.txt

Code structure (Training and Evaluation)

• agents/

  • agent/agents.py contains human/machine action policies, including UCRL2 algorithm.

  • agent/switching_agents.py contains implementation of UCRL2-MC, and UCRL2 for switching agents.

• environments/ contains the code to produce the episodic MDP for the lane driving environment used in the paper.

• experiments/ contains the single team and multiple teams experiment.

Results

See the notebook plots.ipynb to reproduce the results in the paper:

jupyter notebook plots.ipynb