def _assemble_reward(self): self._rewards = [ ir.EvadeUnderAttackRwd(), ir.EvadeInTargetRangeRwd(self._compress_width, self._range_width), sr.ViewEnemyReward(weight=30), ir.EvadeFinalRwd() ]
def _assemble_reward(self): self._explore_rewards = [ ir.EvadeInTargetRangeRwd(self._compress_width, self._range_width), sr.ViewEnemyReward(weight=30) ] self._backward_rewards = [br.BackwardRwd(), br.BackwardFinalRwd()] self._final_rewards = [ir.EvadeUnderAttackRwd(), ir.EvadeFinalRwd()]
def _assemble_reward(self): self._rewards = [ ir.EvadeUnderAttackRwd(), ir.EvadeFinalRwd(), ir.EvadeInTargetRangeRwd(), sr.ViewEnemyReward(weight=20) ]
def _assemble_reward(self): self._rewards = [ ir.EvadeUnderAttackRwd(), fr.FullGameInTargetRangeRwd(self._target_range), fr.FullGameViewRwd(weight=5), ir.EvadeFinalRwd() ]
def _assemble_reward(self): self._rewards = [ ir.EvadeUnderAttackRwd(), ir.EvadeInTargetRangeRwd(self._compress_width, self._range_width), fr.FullGameViewRwd(weight=5), ir.EvadeFinalRwd() ]
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()]
def _assemble_reward(self): self._target = DestRange(self.env.unwrapped.enemy_base(), dest_range=self._target_range) self._first_rewards = [fr.FullGameMoveToTarget(self._target_range)] self._second_rewards = [ ir.EvadeUnderAttackRwd(), fr.FullGameInTargetRangeRwd(self._target_range), fr.FullGameViewRwdV1() ] self._final_rewards = [fr.FullGameFinalRwd()]
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) ]