def run(self, num_episodes, speed):
"""对战调度程序"""
# 启动仿真环境, 与服务端建立rpyc连接
self._start_env()
self.env_client = connect_loop(self.env_manager.get_server_port())
self.env_client.take_action(
[EnvCmd.make_simulation("SPEED", "", speed)])
f = open("state.json", "w")
battle_results = [0, 0, 0] # [红方获胜局数, 平局数量, 蓝方获胜局数]
for i in range(num_episodes):
num_frames = 0
self._reset()
self._run_env()
while True:
try:
num_frames += 1
observation = self._get_observation() # 获取态势
print(i + 1, observation['sim_time'])
print_info(observation['red']['units'])
print()
done = self._get_done(observation) # 推演结束(分出胜负或达到最大时长)
if len(observation['red']['rockets']) > 0:
# 写入所得的json会在同一行里, 打开文件后按Ctrl+Alt+L可自动转换成字典格式
f.write(json.dumps(observation, ensure_ascii=False))
self._run_agents(observation) # 发送指令
if done[0]: # 对战结束后环境重置
# 统计胜负结果
if done[1] == 0 or done[2] == 0:
battle_results[1] += 1
if done[1] == 1:
battle_results[0] += 1
if done[2] == 1:
battle_results[2] += 1
# 环境重置
self.env_manager.reset()
self.env_client = connect_loop(
self.env_manager.get_server_port())
self.env_client.take_action(
[EnvCmd.make_simulation("SPEED", "", speed)])
break
self._run_env()
except Exception as e:
print(e)
print("容器运行出现异常需要重启")
self._start_env()
self.env_client = connect_loop(
self.env_manager.get_server_port())
self.env_client.take_action(
[EnvCmd.make_simulation("SPEED", "", speed)])
break
# 关闭文件
f.close()
return battle_results
def _run_env(self, request_period=5):
new_time = self.env_client.get_time()
self.env_client.take_action([EnvCmd.make_simulation("RESUME", "", "")])
while new_time - self.last_time < request_period:
time.sleep(0.01)
new_time = self.env_client.get_time()
self.last_time = new_time
self.env_client.take_action(
[EnvCmd.make_simulation("PAUSE", "", new_time)])
def _takeoff_areahunt(num, area_hunt_point):
print("红方-轰炸机 执行-起飞区域突击-指令!起飞数量为: {0}!".format(str(num)))
return [
EnvCmd.make_takeoff_areahunt(RED_AIRPORT_ID, num, 270, 80,
*area_hunt_point,
*[270, 1000, 1000, 160])
]
def _takeoff_areapatrol(num, lx, patrol_point, patrol_params):
print("红方-执行-起飞区域巡逻-指令!起飞数量为: {0}!, 起飞战机类型为: {1}!".format(
str(num), str(lx)))
return [
EnvCmd.make_takeoff_areapatrol(RED_AIRPORT_ID, num, lx,
*patrol_point, *patrol_params)
]
def _make_awcs_mode(self_id, mode):
y_mode = ''
if mode == 0:
y_mode = "红方预警机-对空-探测"
elif mode == 1:
y_mode = "红方预警机-对海-探测"
elif mode == 2:
y_mode = "红方预警机-海空交替-探测"
return [EnvCmd.make_awcs_mode(self_id, mode)]
def reset(self):
while True:
self.env_manager.stop_docker()
# time.sleep(random.randint(0, 20))
self.env_manager.start_docker(self.volume_list)
# time.sleep(10)
success = self.env_client.connect_server()
if success == True:
self.env_client.take_action(
[EnvCmd.make_simulation("SPEED", "", self.sim_speed)])
self.last_time = 0
break
self._run_env()
ob = self.env_client.get_observation()
return ob
def _ship_areapatrol(self, self_id, point):
return [
EnvCmd.make_ship_areapatrol(self_id, *point, 180, 20000, 2000, 8,
14400, 0)
]
def _takeoff_targethunt(num, target_id, takeoff_targethunt_params):
return [
EnvCmd.make_takeoff_targethunt(BLUE_AIRPORT_ID, num, target_id,
*takeoff_targethunt_params)
]
def _awacs_patrol(self_id, AWACS_PATROL_POINT, AWACS_PATROL_PARAMS):
return [
EnvCmd.make_awcs_areapatrol(self_id, *AWACS_PATROL_POINT,
*AWACS_PATROL_PARAMS)
]
def _airattack(unit_id, target_id):
return [EnvCmd.make_airattack(unit_id, target_id, 1)]
def _takeoff_areapatrol(num, lx, patrol_point, patrol_params):
return [
EnvCmd.make_takeoff_areapatrol(BLUE_AIRPORT_ID, num, lx,
*patrol_point, *patrol_params)
]
def _ship_areapatrol(self, self_id, point):
return [EnvCmd.make_ship_areapatrol(self_id, *point, *SHIP_PATROL_PARAMS_0)]
def run(self, num_episodes, speed):
global blue_obs
global red_obs
global obs_time
global actions_blue
global actions_red
global dict_actionToUnit
global need_pause
"""对战调度程序"""
# 启动仿真环境, 与服务端建立rpyc连接
self._start_env()
self.env_client = connect_loop(self.env_manager.get_server_port())
self.env_client.take_action([EnvCmd.make_simulation("SPEED", "", speed)])
f = open("state.json", "w")
battle_results = [0, 0, 0] # [红方获胜局数, 平局数量, 蓝方获胜局数]
for i in range(num_episodes):
num_frames = 0
sim_time = 0 # 记录容器或者程序崩溃时的仿真时间
self._reset()
self._run_env()
# 开启记录本轮数据的两个线程
if self.save_replay:
data_port = self.env_manager.get_data_port()
folder = self.replay_dir
net_todocker_init('127.0.0.1', data_port, self.agents[0].name, self.agents[1].name, i, folder)
# red_last=[]
# rockets_lists={}
# f = True
while True:
try:
num_frames += 1
observation = self._get_observation() # 获取态势
sim_time = observation['sim_time']
# print(i + 1, sim_time)
# 获取各种推测态势信息
obs_time = sim_time
# f = True
# if sim_time>=300 and sim_time<=310 and f:
# f = False
# print(sim_time)
# # self.env_manager.stop()
# self.env_client.take_action([EnvCmd.make_simulation('PAUSE','', 0)])
# time.sleep(10)
# print('10秒后')
# self.env_client.take_action([EnvCmd.make_simulation('PAUSE', '', -1)])
if need_pause:
self.env_client.take_action([EnvCmd.make_simulation('PAUSE', '', 0)])
red_obs = observation['red']
blue_obs = observation['blue']
actions_red = self._get_actions_red()
actions_blue = self._get_actions_blue()
caculate_TaskObject(red_obs) #更新任务-对象字典
# rockets_lists = rockets_hitRate(red_obs, red_last, rockets_lists)
print_info(observation['red']['units'])
done = self._get_done(observation) # 推演结束(分出胜负或达到最大时长)
if len(observation['red']['rockets']) > 0:
# 写入所得的json会在同一行里, 打开文件后按Ctrl+Alt+L可自动转换成字典格式
f.write(json.dumps(observation, ensure_ascii=False))
self._run_agents(observation) # 发送指令
if done[0]: # 对战结束后环境重置
# 统计胜负结果
if done[1] == 0 or done[2] == 0:
battle_results[1] += 1
if done[1] == 1:
battle_results[0] += 1
if done[2] == 1:
battle_results[2] += 1
# 环境重置
self.env_manager.reset()
self.env_client = connect_loop(self.env_manager.get_server_port())
self.env_client.take_action([EnvCmd.make_simulation("SPEED", "", speed)])
break
self._run_env()
except Exception as e:
print(e)
print("容器运行出现异常需要重启")
# 记录错误信息
self.logger.debug("第{}局程序运行{}秒后出错, 错误信息为: {}\n".format(i + 1, sim_time, e))
self._start_env()
self.env_client = connect_loop(self.env_manager.get_server_port())
self.env_client.take_action([EnvCmd.make_simulation("SPEED", "", speed)])
break
# 关闭文件
f.close()
return battle_results
def _ship_areapatrol(self_id, point):
print("红方-护卫舰-区域巡逻防空!")
return [
EnvCmd.make_ship_areapatrol(self_id, *point, *SHIP_PATROL_PARAMS_0)
]
def _disturb_stop(self_id):
return [EnvCmd.make_disturb_stop(self_id)]
def _targethunt(self_id, target_id):
return [EnvCmd.make_targethunt(self_id, target_id, 270, 80)]
def _areahunt(self_id, point):
return [EnvCmd.make_areahunt(self_id, 270, 80, *point, *AREA_HUNT_PARAMS)]
def _uav_areapatrol(uav_id, uav_point, uav_params):
return [EnvCmd.make_uav_areapatrol(uav_id, *uav_point, *uav_params)]
def _ship_movedeploy(self, self_id, point):
return [EnvCmd.make_ship_movedeploy(self_id, *point, 90, 1)]
def _awacs_escort(self, awacs_team_id):
return [EnvCmd.make_takeoff_protect(RED_AIRPORT_ID, 8, awacs_team_id, 0, 100, 250)]
def _areapatrol(unit_id, patrol_point, patrol_params):
return [EnvCmd.make_areapatrol(unit_id, *patrol_point, *patrol_params)]
def _disturb_patrol(self, disturb_team_id, patrol_point, patrol_params):
return [EnvCmd.make_disturb_areapatrol(disturb_team_id, *patrol_point, *patrol_params)]
def _returntobase(unit_id):
return [EnvCmd.make_returntobase(unit_id, 20001)]
def _disturb_linepatrol(self, self_id, point_list):
return [EnvCmd.make_disturb_linepatrol(self_id, 160, 0, 'line', point_list)]
def _takeoff_areahunt(num, area_hunt_point, area_hunt_area):
return [
EnvCmd.make_takeoff_areahunt(20001, num, 90, 100, *area_hunt_point,
*area_hunt_area)
]
def _takeoff_areahunt(num, area_hunt_point):
return [EnvCmd.make_takeoff_areahunt(RED_AIRPORT_ID, num, 270, 80, *area_hunt_point, *[270, 1000, 1000, 160])]
def _A2G_escort(self, a2g_team_id):
return [
EnvCmd.make_takeoff_protect(BLUE_AIRPORT_ID, 2, a2g_team_id, 1,
100, 250)
]
def _disturb_escort(self, disturb_team_id):
return [EnvCmd.make_takeoff_protect(RED_AIRPORT_ID, 2, disturb_team_id, 1, 100, 250)]
def run(self, num_episodes, speed):
"""对战调度程序"""
# 启动仿真环境, 与服务端建立rpyc连接
self._start_env()
self.env_client = connect_loop(self.env_manager.get_server_port())
self.env_client.take_action(
[EnvCmd.make_simulation("SPEED", "", speed)])
f = open("state.json", "w")
battle_results = [0, 0, 0] # [红方获胜局数, 平局数量, 蓝方获胜局数]
for i in range(num_episodes):
num_frames = 0
sim_time = 0 # 记录容器或者程序崩溃时的仿真时间
self._reset()
self._run_env()
# 开启记录本轮数据的两个线程
if self.save_replay:
data_port = self.env_manager.get_data_port()
folder = self.replay_dir
net_todocker_init('127.0.0.1', data_port, self.agents[0].name,
self.agents[1].name, i, folder)
while True:
try:
num_frames += 1
observation = self._get_observation() # 获取态势
sim_time = observation['sim_time']
print(i + 1, sim_time)
print_info(observation['red']['units'])
print()
done = self._get_done(observation) # 推演结束(分出胜负或达到最大时长)
if len(observation['red']['rockets']) > 0:
# 写入所得的json会在同一行里, 打开文件后按Ctrl+Alt+L可自动转换成字典格式
f.write(json.dumps(observation, ensure_ascii=False))
self._run_agents(observation) # 发送指令
if done[0]: # 对战结束后环境重置
# 统计胜负结果
if done[1] == 0 or done[2] == 0:
battle_results[1] += 1
if done[1] == 1:
battle_results[0] += 1
if done[2] == 1:
battle_results[2] += 1
# 环境重置
self.env_manager.reset()
self.env_client = connect_loop(
self.env_manager.get_server_port())
self.env_client.take_action(
[EnvCmd.make_simulation("SPEED", "", speed)])
break
self._run_env()
except Exception as e:
print(e)
print("容器运行出现异常需要重启")
# 记录错误信息
self.logger.debug("第{}局程序运行{}秒后出错, 错误信息为: {}\n".format(
i + 1, sim_time, e))
self._start_env()
self.env_client = connect_loop(
self.env_manager.get_server_port())
self.env_client.take_action(
[EnvCmd.make_simulation("SPEED", "", speed)])
break
# 关闭文件
f.close()
return battle_results
def _takeoff_areapatrol(num, lx, patrol_point, patrol_params):
# patrol_params为5个参数
return [EnvCmd.make_takeoff_areapatrol(RED_AIRPORT_ID, num, lx, *patrol_point, *patrol_params)]