@@ -87,16 +87,24 @@ def step(self, action: torch.Tensor):
8787 # 生成observation
8888 obs = self .gen_observation ()
8989
90- # 计算奖励
90+ # 只有动作都执行完才需要检查游戏是否结束
9191 reward = 0
92- # 如果是第一回合 直接return
93- if self .action_history .qsize () < 1 :
94- return obs , reward , False , {}
92+ w_state = self .win_check ()
93+ if w_state != 0 :
94+ reward += 300 if w_state == 2 else - 300
95+ return obs , reward , True , {}
96+
9597 # 计算上一步的奖励
9698 _dirx = [0 , - 1 , 0 , 1 , 1 , - 1 , 1 , - 1 ]
9799 _diry = [- 1 , 0 , 1 , 0 , 1 , - 1 , - 1 , 1 ]
100+ # 如果是第一回合 直接return
101+ if self .action_history .qsize () < 1 :
102+ return obs , reward , False , {}
98103 last_move = self .action_history .queue [- 1 ]
99104 last_obs = self .obs_history .queue [- 1 ]
105+ # 如果动作为空
106+ if last_move [0 ] < 0 :
107+ return obs , reward , False , {}
100108 # 无效移动扣大分
101109 if last_obs [2 ][last_move [0 ] - 1 ][last_move [1 ] - 1 ] != self ._get_colormark (self .learningbot_color ):
102110 reward -= 100
@@ -126,11 +134,6 @@ def step(self, action: torch.Tensor):
126134 self .obs_history .get ()
127135 self .obs_history .put (self .get_view_of (self .learningbot_color ))
128136
129- # 只有动作都执行完才需要检查游戏是否结束
130- w_state = self .win_check ()
131- if w_state != 0 :
132- reward += 300 if w_state == 2 else - 300
133- return obs , reward , True , {}
134137 return obs , reward , False , {}
135138
136139 def render (self , mode = "human" ):
@@ -142,7 +145,7 @@ def render(self, mode="human"):
142145 if mode == "human" :
143146 print_tensor_map (self .map )
144147 print (f"round: { self .round } )
145- time .sleep (1 )
148+ time .sleep (0.65 )
146149
147150 def execute_actions (self , action : torch .Tensor ):
148151 """
@@ -155,22 +158,35 @@ def execute_actions(self, action: torch.Tensor):
155158 cur_color = self .internal_bots_color [i ]
156159 self .internal_bots [cur_color ].bot_move ()
157160 if not self .actions_now [cur_color ]:
161+ print (f"bot { cur_color } )
158162 continue
159163 cur_action = self .actions_now [cur_color ]
160- print (cur_action )
161- f_amount = int (self .map [0 ][cur_action [0 ]][cur_action [1 ]])
162- if cur_action [4 ] == 1 :
163- mov_troop = math .ceil ((f_amount + 0.5 ) / 2 ) - 1
164- else :
165- mov_troop = f_amount - 1
166- self .combine ((cur_action [0 ], cur_action [1 ]), (cur_action [2 ], cur_action [3 ]), mov_troop )
164+ # 如果移动超界了 直接下一个
165+ skip = False
166+ for j in range (4 ):
167+ if not 0 <= cur_action [j ] < self .map_size :
168+ skip = True
169+ break
170+ if skip :
171+ print (f"skipped: { cur_action } )
172+ continue
173+
174+ print (f"internal bot color { cur_color } { cur_action } )
175+ # 检查动作是否合法
176+ if cur_action [0 ] >= 0 and self .map [2 ][cur_action [0 ]][cur_action [1 ]] == cur_color :
177+ f_amount = int (self .map [0 ][cur_action [0 ]][cur_action [1 ]])
178+ if cur_action [4 ] == 1 :
179+ mov_troop = math .ceil ((f_amount + 0.5 ) / 2 ) - 1
180+ else :
181+ mov_troop = f_amount - 1
182+ self .combine ((cur_action [0 ], cur_action [1 ]), (cur_action [2 ], cur_action [3 ]), mov_troop )
167183
168184 # 处理LearningBot动作
169185 act = self .at .i_to_a (self .map_size , int (action ))[0 ].long ()
170186 act -= 1
171187 act [4 ] += 1
172188 # 检查动作是否合法
173- if self .map [2 ][act [0 ]][act [1 ]] == self .learningbot_color :
189+ if act [ 0 ] >= 0 and self .map [2 ][act [0 ]][act [1 ]] == self .learningbot_color :
174190 f_amount = int (self .map [0 ][act [0 ]][act [1 ]])
175191 if act [4 ] == 1 :
176192 mov_troop = math .ceil ((f_amount + 0.5 ) / 2 ) - 1
@@ -197,6 +213,10 @@ def combine(self, b1: tuple, b2: tuple, cnt):
197213 "color" : int (self .map [2 ][b2 [0 ]][b2 [1 ]])
198214 }
199215
216+ # 地形特判
217+ if t ["type" ] == BlockType .mountain :
218+ return
219+
200220 if t ["color" ] == f ["color" ]:
201221 t ["amount" ] += cnt
202222 f ["amount" ] -= cnt
@@ -213,7 +233,7 @@ def combine(self, b1: tuple, b2: tuple, cnt):
213233 if self .map [2 ][i ][j ] == BlockType .crown :
214234 self .map [2 ][i ][j ] = BlockType .city
215235 t ["color" ] = f ["color" ]
216- t ["amount" ] = f ["amount" ]
236+ t ["amount" ] = - t ["amount" ]
217237
218238 # 赋值回去
219239 self .map [0 ][b1 [0 ]][b1 [1 ]] = f ["amount" ]
@@ -303,9 +323,13 @@ def bot_action_upd(self, color, action):
303323 :param action: [x1, y1, x2, y2, is_half]
304324 :return:
305325 """
306- print (f"internal bot color { color } { action } )
307326 if action :
308- self .actions_now [color ] = action
327+ self .actions_now [color ] = [None , None , None , None , None ]
328+ self .actions_now [color ][0 ] = action [0 ] - 1
329+ self .actions_now [color ][1 ] = action [1 ] - 1
330+ self .actions_now [color ][2 ] = action [2 ] - 1
331+ self .actions_now [color ][3 ] = action [3 ] - 1
332+ self .actions_now [color ][4 ] = action [4 ]
309333 else :
310334 self .actions_now [color ] = False
311335
0 commit comments