Пример #1
0
 def _assemble_reward(self):
     self._rewards = [
         sr.HomeReward(),
         sr.EnemyBaseReward(),
         sr.ViewEnemyReward(),
         sr.MinDistReward()
     ]
Пример #2
0
    def _assemble_reward(self):
        self._forward_rewards = [
            sr.HomeReward(negative=True),
            sr.EnemyBaseReward(negative=True),
            sr.ViewEnemyReward(weight=10),
            sr.EnemyBaseArrivedReward(weight=30),
            sr.MinDistReward(negative=True)
        ]

        self._backward_rewards = [
            sr.HomeReward(back=True, negative=True),
            sr.EnemyBaseReward(back=True, negative=True),
            sr.HomeArrivedReward(weight=30),
            sr.MinDistReward(negative=True)
        ]

        self._final_rewards = [sr.RoundTripFinalReward(weight=50)]
Пример #3
0
 def _assemble_reward(self):
     self._rewards = [
         sr.HomeReward(negative=True),
         sr.EnemyBaseReward(negative=True),
         sr.ViewEnemyReward(weight=10),
         sr.MinDistReward(negative=True),
         sr.EnemyBaseArrivedReward(weight=30),
         sr.OnewayFinalReward(weight=50)
     ]
Пример #4
0
 def _assemble_reward(self):
     self._forward_rewards = [
         sr.HomeReward(negative=True),
         sr.EnemyBaseReward(negative=True),
         sr.MinDistReward(negative=True)
     ]
     self._backward_rewards = [
         sr.HomeReward(negative=True, back=True),
         sr.EnemyBaseReward(negative=True, back=True),
         ir.EvadeUnderAttackRwd()
     ]
     self._explore_rewards = [
         ir.EvadeUnderAttackRwd(),
         ir.EvadeInTargetRangeRwd(self._compress_width, self._range_width),
         sr.ViewEnemyReward(weight=20)
         #ir.EvadeTargetScanRwd(self._compress_width, self._range_width),
     ]
     self._final_rewards = [ir.EvadeFinalRwd()]
Пример #5
0
    def _assemble_reward(self):
        self._forward_rewards = [
            sr.HomeReward(negative=True, weight=0.1),
            sr.EnemyBaseReward(negative=True, weight=0.1),
            sr.ViewEnemyReward(weight=0.5),
            # sr.EnemyBaseArrivedReward(weight=50),
            sr.MinDistReward(negative=True, weight=0.1),
            # er.EvadeDistanceReward(weight=1),
            er.EvadeUnderAttackRwd(weight=0.5),
            # er.EnemyInRangeRwd(weight=1),
            # sr.AreaOfOverlapReward(weight=2)
        ]

        self._explore_rewards = [
            # sr.HomeReward(negative=True,weight=1),
            # sr.ViewEnemyResourcesAndBase(weight=10),
            sr.ExploreStateRwd(weight=1),
            # er.EvadeDistanceReward(weight=1),
            er.EvadeUnderAttackRwd(weight=0.5),
            sr.ViewEnemyResourcesAndBase(weight=1),
            # er.EnemyInRangeRwd(weight=1),
            sr.ExploreAcclerateRwd(weight=0.05)
        ]

        self._backward_rewards = [
            sr.HomeReward(back=True, negative=True, weight=0.1),
            #sr.EnemyBaseReward(back=True, negative=False,weight=1),
            sr.HomeArrivedReward(weight=1),
            #sr.MinDistReward(negative=True),
            sr.BackwardStateRwd(weight=1),
            # er.EvadeDistanceReward(weight=1),
            er.EvadeUnderAttackRwd(weight=0.5)
            # er.EnemyInRangeRwd(weight=1),
            # sr.AreaOfOverlapReward(weight=2),
            # sr.HitEnemyBaseReward(weight=50),
        ]

        self._final_rewards = [
            sr.RoundTripFinalReward(weight=1),
            er.EvadeFinalRwd(weight=1)
        ]