AWS Deepracer Simulation round solution

Solution was developed in collaboration with @DAOvb.

challenge link

Solution is based on a simple PPO algorithm, achieves near-optimal score in AWS Deepracer environment.

To start training:

docker-compose up -d

To create a gif of agent’s rollout:

docker-compose up -d
python save/best_416_iconic-field-98_6env_SCLight_4st1sk.pickle 

Example gif


View Github