simulation.py

import random
import numpy as np


def simulation(env, attacker, nn_att, defender, nn_def, num_episodes):
    #TODO: APIs have been changed.
    aReward_list = np.array([])
    dReward_list = np.array([])

    for i in range(num_episodes): #can be run parallel
        aReward = 0
        dReward = 0
        env.G.reset()
        attacker.reset_att()
        defender.reset_def()
        for t in range(env.T):
            timeleft = env.T - t
            attacker.att_greedy_action_builder_single(env.G, timeleft, nn_att)
            att_action_set = attacker.attact
            defender.def_greedy_action_builder_single(env.G, timeleft, nn_def)
            def_action_set = defender.defact
            for attack in att_action_set:
                if isinstance(attack, tuple):
                    # check OR node
                    aReward += env.G.edges[attack]['cost']
                    if random.uniform(0, 1) <= env.G.edges[attack]['actProb']:
                        env.G.nodes[attack[-1]]['state'] = 1
                else:
                    # check AND node
                    aReward += env.G.nodes[attack]['aCost']
                    if random.uniform(0, 1) <= env.G.nodes[attack]['actProb']:
                        env.G.nodes[attack]['state'] = 1
            # defender's action
            for node in def_action_set:
                env.G.nodes[node]['state'] = 0
                dReward += env.G.nodes[node]['dCost']
            _, targetset = env.get_Targets()
            for node in targetset:
                if env.G.nodes[node]['state'] == 1:
                    aReward += env.G.nodes[node]['aReward']
                    dReward += env.G.nodes[node]['dPenalty']

        aReward_list = np.append(aReward_list,aReward)
        dReward_list = np.append(dReward_list,dReward)

    return np.mean(aReward_list), np.mean(dReward_list)