DQN本地训练不能持续

[leviJ123]

TTKP 项目DQN训练时根据教程操作，双击start test能够进入游戏但游戏不能持续训练，mc,io等模块自动退出,
并且没有显示result游戏界面（和教程中双击start test弹出result界面存在差异），且运行agent.py有以下报错：
[2021-07-01 16:16:17,156][agentai.py:135][ERROR] : exception: 'ROI' msg: Traceback (most recent call last):
File "agentai.py", line 131, in Main
RunFrameWork(args.mode)
File "agentai.py", line 94, in RunFrameWork
aiFramework.Run(True)
File "../src/AgentAI/aiframework/AIFrameWork.py", line 116, in Run
self._DefaultRun(isTestMode)
File "../src/AgentAI/aiframework/AIFrameWork.py", line 142, in _DefaultRun
self._WaitEpisode()
File "../src/AgentAI/aiframework/AIFrameWork.py", line 161, in _WaitEpisode
if self.__agentEnv.IsEpisodeStart() is True:
File "../src/AgentAI/agentenv/DQNEnv.py", line 171, in IsEpisodeStart
self._GetGameInfo()
File "../src/AgentAI/agentenv/DQNEnv.py", line 224, in _GetGameInfo
self.update_scene_task(result, self.__actionController.get_action_dict(), self.__agentAPI)
File "../src/AgentAI/agentenv/GameEnv.py", line 136, in update_scene_task
flag, px, py = self._get_position(result_dict, scene_task_id)
File "../src/AgentAI/agentenv/GameEnv.py", line 160, in _get_position
x = result['ROI']['x']
KeyError: 'ROI'

请问是哪里的问题
操作系统是Ubuntu16.04