Python3 UnicodeDecodeError: 'utf-8' codec can't decode byte 0xfe in position 63: invalid start byte

前提・実現したいこと

githubから以下のコードをダウンロードして実行したのですが、表題のエラーが発生し実行することができません。
Unicodeに関するエラーなのですが、解決方法を教えていただけないでしょうか。
[Github]ArmRobot

エラーの詳細を記述していきます。
ターミナル上で、上記のコードを実行した時、
"ArmRobot-MLmaster/python/unityagents/environment.py"ファイル内の138行目
"p = self._conn.recv(self._buffer_size).decode('utf-8')"にて下記のようなエラーメッセージが発生します。
デコードができないと怒られているようです。
こちらのエラーの対処法をご存知の方がおられたら教えていただけないでしょうか？

発生している問題・エラーメッセージ

line 138, in __init__
    p = self._conn.recv(self._buffer_size).decode('utf-8')
UnicodeDecodeError: 'utf-8' codec can't decode byte 0xfe in position 63: invalid start byte

該当のソースコード

python
1import atexit
2import io
3import glob
4import json
5import logging
6import numpy as np
7import os
8import socket
9import subprocess
10import struct
11
12from .brain import BrainInfo, BrainParameters, AllBrainInfo
13from .exception import UnityEnvironmentException, UnityActionException, UnityTimeOutException
14from .curriculum import Curriculum
15
16from PIL import Image
17from sys import platform
18
19logging.basicConfig(level=logging.INFO)
20logger = logging.getLogger("unityagents")
21
22
23class UnityEnvironment(object):
24    def __init__(self, file_name, worker_id=0,
25                 base_port=5005, curriculum=None,
26                 seed=0, docker_training=False):
27        """
28        Starts a new unity environment and establishes a connection with the environment.
29        Notice: Currently communication between Unity and Python takes place over an open socket without authentication.
30        Ensure that the network where training takes place is secure.
31
32        :string file_name: Name of Unity environment binary.
33        :int base_port: Baseline port number to connect to Unity environment over. worker_id increments over this.
34        :int worker_id: Number to add to communication port (5005) [0]. Used for asynchronous agent scenarios.
35        :param docker_training: Informs this class whether the process is being run within a container.
36        """
37
38        atexit.register(self.close)
39        self.port = base_port + worker_id
40        self._buffer_size = 12000
41        self._version_ = "API-3"
42        self._loaded = False
43        self._open_socket = False
44
45        try:
46            # Establish communication socket
47            self._socket = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
48            self._socket.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
49            self._socket.bind(("localhost", self.port))
50            self._open_socket = True
51        except socket.error:
52            self._open_socket = True
53            self.close()
54            raise socket.error("Couldn't launch new environment because worker number {} is still in use. "
55                               "You may need to manually close a previously opened environment "
56                               "or use a different worker number.".format(str(worker_id)))
57
58        cwd = os.getcwd()
59        file_name = (file_name.strip()
60                     .replace('.app', '').replace('.exe', '').replace('.x86_64', '').replace('.x86', ''))
61        true_filename = os.path.basename(os.path.normpath(file_name))
62        logger.debug('The true file name is {}'.format(true_filename))
63        launch_string = None
64        if platform == "linux" or platform == "linux2":
65            candidates = glob.glob(os.path.join(cwd, file_name) + '.x86_64')
66            if len(candidates) == 0:
67                candidates = glob.glob(os.path.join(cwd, file_name) + '.x86')
68            if len(candidates) == 0:
69                candidates = glob.glob(file_name + '.x86_64')
70            if len(candidates) == 0:
71                candidates = glob.glob(file_name + '.x86')
72            if len(candidates) > 0:
73                launch_string = candidates[0]
74
75        elif platform == 'darwin':
76            candidates = glob.glob(os.path.join(cwd, file_name + '.app', 'Contents', 'MacOS', true_filename))
77            if len(candidates) == 0:
78                candidates = glob.glob(os.path.join(file_name + '.app', 'Contents', 'MacOS', true_filename))
79            if len(candidates) == 0:
80                candidates = glob.glob(os.path.join(cwd, file_name + '.app', 'Contents', 'MacOS', '*'))
81            if len(candidates) == 0:
82                candidates = glob.glob(os.path.join(file_name + '.app', 'Contents', 'MacOS', '*'))
83            if len(candidates) > 0:
84                launch_string = candidates[0]
85        elif platform == 'win32':
86            candidates = glob.glob(os.path.join(cwd, file_name + '.exe'))
87            if len(candidates) == 0:
88                candidates = glob.glob(file_name + '.exe')
89            if len(candidates) > 0:
90                launch_string = candidates[0]
91        if launch_string is None:
92            self.close()
93            raise UnityEnvironmentException("Couldn't launch the {0} environment. "
94                                            "Provided filename does not match any environments."
95                                            .format(true_filename))
96        else:
97            logger.debug("This is the launch string {}".format(launch_string))
98            # Launch Unity environment
99            if docker_training == False:
100                proc1 = subprocess.Popen(
101                    [launch_string,
102                     '--port', str(self.port),
103                     '--seed', str(seed)])
104            else:
105                """
106                Comments for future maintenance:
107                    xvfb-run is a wrapper around Xvfb, a virtual xserver where all
108                    rendering is done to virtual memory. It automatically creates a
109                    new virtual server automatically picking a server number `auto-servernum`.
110                    The server is passed the arguments using `server-args`, we are telling
111                    Xvfb to create Screen number 0 with width 640, height 480 and depth 24 bits.
112                    Note that 640 X 480 are the default width and height. The main reason for
113                    us to add this is because we'd like to change the depth from the default
114                    of 8 bits to 24.
115                    Unfortunately, this means that we will need to pass the arguments through
116                    a shell which is why we set `shell=True`. Now, this adds its own
117                    complications. E.g SIGINT can bounce off the shell and not get propagated
118                    to the child processes. This is why we add `exec`, so that the shell gets
119                    launched, the arguments are passed to `xvfb-run`. `exec` replaces the shell
120                    we created with `xvfb`.
121                """
122                docker_ls = ("exec xvfb-run --auto-servernum"
123                             " --server-args='-screen 0 640x480x24'"
124                             " {0} --port {1} --seed {2}").format(launch_string,
125                                                                  str(self.port),
126                                                                  str(seed))
127                proc1 = subprocess.Popen(docker_ls,
128                                         stdout=subprocess.PIPE,
129                                         stderr=subprocess.PIPE,
130                                         shell=True)
131        self._socket.settimeout(30)
132        try:
133            try:
134                self._socket.listen(1)
135                self._conn, _ = self._socket.accept()
136                self._conn.settimeout(30)
137                import pdb; pdb.set_trace()
138#下記の行でエラーが発生します
139                p = self._conn.recv(self._buffer_size).decode('utf-8')
140                p = json.loads(p)
141            except socket.timeout as e:
142                raise UnityTimeOutException(
143                    "The Unity environment took too long to respond. Make sure {} does not need user interaction to "
144                    "launch and that the Academy and the external Brain(s) are attached to objects in the Scene."
145                    .format(str(file_name)))
146#teratailの仕様上10000文字以上の投稿はできないようなので、これ以下のコードは省略します。お手数ですが、上記のリンクのgithubから参照ください
147

試したこと

・実行したシェルの文字コードの確認(echo $LANG)
ターミナルの文字コードはja_JP.UTF-8
ターミナルとコードの文字コードの不一致によるエラーの発生の線は薄そうです。

・デバッグ(pdb)
変数pには何も格納されていませんでした。
socketで送られてきたデータの中身を確認することは叶わず

・インターネットで調べてみるも解決方法に至るような情報は見つからず
python3でのbytes型の中で.decode('utf-8')を指定してもエラーになる
 技術系メモ
 UnicodeDecodeError: 'utf-8' codec can't decode byte 0xbc in position 0: invalid start byte

補足情報（FW/ツールのバージョンなど）

macOS Sierra 10.12.6
Unity 2018.2.4.f1
ターミナル2.7.3(文字コード ja_JP.UTF-8 )
Python 3.6.6 :: Anaconda, Inc.

行動規範の内容に同意します

回答2件

ベストアンサー

UnicodeDecodeError: 'utf-8' codec can't decode byte 0xfe

~~憶測ですがdecode byte 0xfeなので、utf-8ではなくutf-16でデータが送信されているように見えます。~~

Python
1    def __init__(self, file_name, worker_id=0,
2                 base_port=5005, curriculum=None,
3                 seed=0, docker_training=False):