We will evaluate your system on 5 maps. Copy the maps into the map folder of your gym-duckietown environment, e.g.,
cp maps/* /path/to/your-gym-duckietown-repo/gym-duckietown/maps/
Each map is associated with several random seeds.
To initialize the gym-duckietown environment, there are three arguments:
--map-name
: the name of the map--seed
: random seed of the environment--max-steps
: the maximum run step. The default value is 2000. Do not change this default value when you generate the control files for submission.
Similar to Assignment 3, you generate the control files for submission.
A sample file for the python environment is available here. We also include a sample control file. It illustrates how to add arguments and output your controls to a file. To try our simple policy,
python example.py --map-name map5 --seed 11
We will evaluate your system on these 5 maps and compute the accumulated reward for each test case. A primary component of your grade is the average reward achieved.