Adaptive Cruise Control for a Hybrid Vehicle with Deep Policy Gradients

Final project for ECE 517/414 Reinforcement Learning.

Running Code

Example code is provided to train the REINFORCE and DDPG algorithms and view training curves over a small number of steps. This requires the confidential Blazer Model, and it cannot be ran without it. Each test script is ran with the raw speed reward function.

REINFORCE

Run the following.

python3 ACC_RL/test_scripts/test_REINFORCE.py

DDPG

Run the following. Be cautioned that it takes ~2 minutes and will output a plot.

python3 ACC_RL/test_scripts/test_DDPG.py