Apply reinforcement learning to an agent in a 13x13 four-room bounded environment. Runs under the following scenarios:
- 1: Single package
- 2: Multiple packages
- 3: Multiple packages with ordered collection
Run make
inside the extracted folder. Pip will install its dependencies inside a virtual environment
usage: ExecutionSkeleton [-h] [-stochastic] [-learning-rate LEARNING_RATE] [-discount-rate DISCOUNT_RATE]
[-epochs EPOCHS] [-test] [-save SAVE]
scenario
Trains the FourRooms Agent in a given scenario
positional arguments:
scenario
options:
-h, --help show this help message and exit
-stochastic
-learning-rate LEARNING_RATE
-discount-rate DISCOUNT_RATE
-epochs EPOCHS
-test
-save SAVE
FourRooms.py
: Environment frameworkExecutionSkeleton.py
: Agent definition and runnersScenario1.py
: Run thesimple
scenarioScenario2.py
: Run themulti
scenarioScenario3.py
: Run thergb
scenario