The Predator Prey Project

The rl-predator-prey code is a reconfiguration of https://github.com/andrearama/PredatorPreyReinfLearning. The primary goal is:

• to reproduce this algorithm in a PettingZoo environment with an Agent Environment Cycle (AEC)

Tweaks

• After eating, Prey as well as Predators get a chance of reproduction. Changed because in the original code an abundance of food does not lead to reproduction opportunities. A discrepancy exists in the author's article(2015) and presentation(2014), see flow charts below.

• In the method update_weight, the last feature in the Moore environment was erroneously used to update the weights. Has been corrected tot the feature of the current location of the Predator or Prey in predator.py and prey.py

• An attempt was made to reconfigure the code towards standardization in a gym fashioned way. The state space array has been brought back to one object (rather than three separate arrays for each agent).

Diagrams