Collaboration by the tigers
The second experiment that I implemented was designed to make the tigers’ lives more complicated and encourage collaboration between them. The training and play code are the same; the only difference is in the MAgent environment’s configuration.
If you pass the argument --mode double_attack to the training utility, the environment data.DoubleAttackEnv will be used. The only difference is the configuration object, which sets additional constraints on tigers’ attacks. In the new setup, they can attack deer only in pairs and have to do this at the same time. A single tiger’s attack doesn’t have any effect. This definitely complicates the training and hunting, as obtaining the reward from eating the deer is now much harder for tigers. To start the training, you can run the same train utility, but with an extra command-line argument:
./forest_tigers_dqn.py -n run-name --dev cuda --mode double_attackLet...
