Windows下OpenAI gym環境的使用
作者：凱魯嘎吉 - 博客園 http://www.cnblogs.com/kailugaji/
1. gym環境搭建用到的關鍵語句
1.1 准備工作
首先創建一個虛擬環境conda create -n RL python=3.8，激活activate RL。我用到的包及版本conda list:
ale-py 0.7.3 <pip>
atari-py 1.2.2 <pip>
Box2D 2.3.10 <pip>
box2d-py 2.3.8 <pip>
ca-certificates 2021.10.26 haa95532_2 http://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main
certifi 2020.6.20 py37_0 anaconda
cffi 1.15.0 <pip>
cloudpickle 2.0.0 <pip>
cycler 0.11.0 <pip>
Cython 0.29.26 <pip>
fasteners 0.16.3 <pip>
ffmpeg 1.4 <pip>
fonttools 4.28.5 <pip>
glfw 2.5.0 <pip>
gym 0.21.0 <pip>
imageio 2.13.5 <pip>
importlib-metadata 2.0.0 py_1 anaconda
importlib-resources 5.4.0 <pip>
kiwisolver 1.3.2 <pip>
lockfile 0.12.2 <pip>
matplotlib 3.5.1 <pip>
mujoco-py 1.50.1.68 <pip>
numpy 1.21.5 <pip>
openssl 1.0.2t vc14h62dcd97_0 [vc14] anaconda
packaging 21.3 <pip>
Pillow 9.0.0 <pip>
pip 20.2.4 py37_0 anaconda
pycparser 2.21 <pip>
pyglet 1.5.21 <pip>
pyparsing 3.0.6 <pip>
python 3.7.1 h33f27b4_4 anaconda
python-dateutil 2.8.2 <pip>
setuptools 50.3.0 py37h9490d1a_1 anaconda
six 1.16.0 <pip>
sqlite 3.20.1 vc14h7ce8c62_1 [vc14] anaconda
swig 3.0.12 h047fa9f_3 anaconda
vc 14.2 h21ff451_1 http://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main
vs2015_runtime 14.27.29016 h5e58377_2 http://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main
wheel 0.37.0 pyhd3eb1b0_1 http://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main
wincertstore 0.2 py37_0 anaconda
wrappers 0.1.9 <pip>
zipp 3.3.1 py_0 anaconda
zipp 3.7.0 <pip>
然后安裝numpy: pip install numpy
1.2 安裝gym, atari, Box2D, mujoco
1.2.1 安裝簡單版的gym
pip install gym, pip install pyglet
查看所有模擬環境:
from gym import envs
names = [env.id for env in envs.registry.all()]
print('\n'.join(names))
ALE/Tetris-v5 ALE/Tetris-ram-v5 ALE/Adventure-v5 ALE/Adventure-ram-v5 ALE/AirRaid-v5 ALE/AirRaid-ram-v5 ALE/Alien-v5 ALE/Alien-ram-v5 ALE/Asterix-v5 ALE/Asterix-ram-v5 ALE/Asteroids-v5 ALE/Asteroids-ram-v5 ALE/BeamRider-v5 ALE/BeamRider-ram-v5 ALE/Bowling-v5 ALE/Bowling-ram-v5 ALE/Boxing-v5 ALE/Boxing-ram-v5 ALE/Breakout-v5 ALE/Breakout-ram-v5 ALE/ChopperCommand-v5 ALE/ChopperCommand-ram-v5 ALE/CrazyClimber-v5 ALE/CrazyClimber-ram-v5 ALE/ElevatorAction-v5 ALE/ElevatorAction-ram-v5 ALE/Enduro-v5 ALE/Enduro-ram-v5 ALE/FishingDerby-v5 ALE/FishingDerby-ram-v5 ALE/Freeway-v5 ALE/Freeway-ram-v5 ALE/Frostbite-v5 ALE/Frostbite-ram-v5 ALE/Hero-v5 ALE/Hero-ram-v5 ALE/Jamesbond-v5 ALE/Jamesbond-ram-v5 ALE/JourneyEscape-v5 ALE/JourneyEscape-ram-v5 ALE/Kangaroo-v5 ALE/Kangaroo-ram-v5 ALE/MontezumaRevenge-v5 ALE/MontezumaRevenge-ram-v5 ALE/Phoenix-v5 ALE/Phoenix-ram-v5 ALE/Pong-v5 ALE/Pong-ram-v5 ALE/PrivateEye-v5 ALE/PrivateEye-ram-v5 ALE/Qbert-v5 ALE/Qbert-ram-v5 ALE/Robotank-v5 ALE/Robotank-ram-v5 ALE/Seaquest-v5 ALE/Seaquest-ram-v5 ALE/Solaris-v5 ALE/Solaris-ram-v5 ALE/StarGunner-v5 ALE/StarGunner-ram-v5 ALE/TimePilot-v5 ALE/TimePilot-ram-v5 ALE/UpNDown-v5 ALE/UpNDown-ram-v5 ALE/Zaxxon-v5 ALE/Zaxxon-ram-v5 Adventure-v0 AdventureDeterministic-v0 AdventureNoFrameskip-v0 Adventure-v4 AdventureDeterministic-v4 AdventureNoFrameskip-v4 Adventure-ram-v0 Adventure-ramDeterministic-v0 Adventure-ramNoFrameskip-v0 Adventure-ram-v4 Adventure-ramDeterministic-v4 Adventure-ramNoFrameskip-v4 AirRaid-v0 AirRaidDeterministic-v0 AirRaidNoFrameskip-v0 AirRaid-v4 AirRaidDeterministic-v4 AirRaidNoFrameskip-v4 AirRaid-ram-v0 AirRaid-ramDeterministic-v0 AirRaid-ramNoFrameskip-v0 AirRaid-ram-v4 AirRaid-ramDeterministic-v4 AirRaid-ramNoFrameskip-v4 Alien-v0 AlienDeterministic-v0 AlienNoFrameskip-v0 Alien-v4 AlienDeterministic-v4 AlienNoFrameskip-v4 Alien-ram-v0 Alien-ramDeterministic-v0 Alien-ramNoFrameskip-v0 Alien-ram-v4 Alien-ramDeterministic-v4 Alien-ramNoFrameskip-v4 Amidar-v0 AmidarDeterministic-v0 AmidarNoFrameskip-v0 Amidar-v4 AmidarDeterministic-v4 AmidarNoFrameskip-v4 Amidar-ram-v0 Amidar-ramDeterministic-v0 Amidar-ramNoFrameskip-v0 Amidar-ram-v4 Amidar-ramDeterministic-v4 Amidar-ramNoFrameskip-v4 Assault-v0 AssaultDeterministic-v0 AssaultNoFrameskip-v0 Assault-v4 AssaultDeterministic-v4 AssaultNoFrameskip-v4 Assault-ram-v0 Assault-ramDeterministic-v0 Assault-ramNoFrameskip-v0 Assault-ram-v4 Assault-ramDeterministic-v4 Assault-ramNoFrameskip-v4 Asterix-v0 AsterixDeterministic-v0 AsterixNoFrameskip-v0 Asterix-v4 AsterixDeterministic-v4 AsterixNoFrameskip-v4 Asterix-ram-v0 Asterix-ramDeterministic-v0 Asterix-ramNoFrameskip-v0 Asterix-ram-v4 Asterix-ramDeterministic-v4 Asterix-ramNoFrameskip-v4 Asteroids-v0 AsteroidsDeterministic-v0 AsteroidsNoFrameskip-v0 Asteroids-v4 AsteroidsDeterministic-v4 AsteroidsNoFrameskip-v4 Asteroids-ram-v0 Asteroids-ramDeterministic-v0 Asteroids-ramNoFrameskip-v0 Asteroids-ram-v4 Asteroids-ramDeterministic-v4 Asteroids-ramNoFrameskip-v4 Atlantis-v0 AtlantisDeterministic-v0 AtlantisNoFrameskip-v0 Atlantis-v4 AtlantisDeterministic-v4 AtlantisNoFrameskip-v4 Atlantis-ram-v0 Atlantis-ramDeterministic-v0 Atlantis-ramNoFrameskip-v0 Atlantis-ram-v4 Atlantis-ramDeterministic-v4 Atlantis-ramNoFrameskip-v4 BankHeist-v0 BankHeistDeterministic-v0 BankHeistNoFrameskip-v0 BankHeist-v4 BankHeistDeterministic-v4 BankHeistNoFrameskip-v4 BankHeist-ram-v0 BankHeist-ramDeterministic-v0 BankHeist-ramNoFrameskip-v0 BankHeist-ram-v4 BankHeist-ramDeterministic-v4 BankHeist-ramNoFrameskip-v4 BattleZone-v0 BattleZoneDeterministic-v0 BattleZoneNoFrameskip-v0 BattleZone-v4 BattleZoneDeterministic-v4 BattleZoneNoFrameskip-v4 BattleZone-ram-v0 BattleZone-ramDeterministic-v0 BattleZone-ramNoFrameskip-v0 BattleZone-ram-v4 BattleZone-ramDeterministic-v4 BattleZone-ramNoFrameskip-v4 BeamRider-v0 BeamRiderDeterministic-v0 BeamRiderNoFrameskip-v0 BeamRider-v4 BeamRiderDeterministic-v4 BeamRiderNoFrameskip-v4 BeamRider-ram-v0 BeamRider-ramDeterministic-v0 BeamRider-ramNoFrameskip-v0 BeamRider-ram-v4 BeamRider-ramDeterministic-v4 BeamRider-ramNoFrameskip-v4 Berzerk-v0 BerzerkDeterministic-v0 BerzerkNoFrameskip-v0 Berzerk-v4 BerzerkDeterministic-v4 BerzerkNoFrameskip-v4 Berzerk-ram-v0 Berzerk-ramDeterministic-v0 Berzerk-ramNoFrameskip-v0 Berzerk-ram-v4 Berzerk-ramDeterministic-v4 Berzerk-ramNoFrameskip-v4 Bowling-v0 BowlingDeterministic-v0 BowlingNoFrameskip-v0 Bowling-v4 BowlingDeterministic-v4 BowlingNoFrameskip-v4 Bowling-ram-v0 Bowling-ramDeterministic-v0 Bowling-ramNoFrameskip-v0 Bowling-ram-v4 Bowling-ramDeterministic-v4 Bowling-ramNoFrameskip-v4 Boxing-v0 BoxingDeterministic-v0 BoxingNoFrameskip-v0 Boxing-v4 BoxingDeterministic-v4 BoxingNoFrameskip-v4 Boxing-ram-v0 Boxing-ramDeterministic-v0 Boxing-ramNoFrameskip-v0 Boxing-ram-v4 Boxing-ramDeterministic-v4 Boxing-ramNoFrameskip-v4 Breakout-v0 BreakoutDeterministic-v0 BreakoutNoFrameskip-v0 Breakout-v4 BreakoutDeterministic-v4 BreakoutNoFrameskip-v4 Breakout-ram-v0 Breakout-ramDeterministic-v0 Breakout-ramNoFrameskip-v0 Breakout-ram-v4 Breakout-ramDeterministic-v4 Breakout-ramNoFrameskip-v4 Carnival-v0 CarnivalDeterministic-v0 CarnivalNoFrameskip-v0 Carnival-v4 CarnivalDeterministic-v4 CarnivalNoFrameskip-v4 Carnival-ram-v0 Carnival-ramDeterministic-v0 Carnival-ramNoFrameskip-v0 Carnival-ram-v4 Carnival-ramDeterministic-v4 Carnival-ramNoFrameskip-v4 Centipede-v0 CentipedeDeterministic-v0 CentipedeNoFrameskip-v0 Centipede-v4 CentipedeDeterministic-v4 CentipedeNoFrameskip-v4 Centipede-ram-v0 Centipede-ramDeterministic-v0 Centipede-ramNoFrameskip-v0 Centipede-ram-v4 Centipede-ramDeterministic-v4 Centipede-ramNoFrameskip-v4 ChopperCommand-v0 ChopperCommandDeterministic-v0 ChopperCommandNoFrameskip-v0 ChopperCommand-v4 ChopperCommandDeterministic-v4 ChopperCommandNoFrameskip-v4 ChopperCommand-ram-v0 ChopperCommand-ramDeterministic-v0 ChopperCommand-ramNoFrameskip-v0 ChopperCommand-ram-v4 ChopperCommand-ramDeterministic-v4 ChopperCommand-ramNoFrameskip-v4 CrazyClimber-v0 CrazyClimberDeterministic-v0 CrazyClimberNoFrameskip-v0 CrazyClimber-v4 CrazyClimberDeterministic-v4 CrazyClimberNoFrameskip-v4 CrazyClimber-ram-v0 CrazyClimber-ramDeterministic-v0 CrazyClimber-ramNoFrameskip-v0 CrazyClimber-ram-v4 CrazyClimber-ramDeterministic-v4 CrazyClimber-ramNoFrameskip-v4 Defender-v0 DefenderDeterministic-v0 DefenderNoFrameskip-v0 Defender-v4 DefenderDeterministic-v4 DefenderNoFrameskip-v4 Defender-ram-v0 Defender-ramDeterministic-v0 Defender-ramNoFrameskip-v0 Defender-ram-v4 Defender-ramDeterministic-v4 Defender-ramNoFrameskip-v4 DemonAttack-v0 DemonAttackDeterministic-v0 DemonAttackNoFrameskip-v0 DemonAttack-v4 DemonAttackDeterministic-v4 DemonAttackNoFrameskip-v4 DemonAttack-ram-v0 DemonAttack-ramDeterministic-v0 DemonAttack-ramNoFrameskip-v0 DemonAttack-ram-v4 DemonAttack-ramDeterministic-v4 DemonAttack-ramNoFrameskip-v4 DoubleDunk-v0 DoubleDunkDeterministic-v0 DoubleDunkNoFrameskip-v0 DoubleDunk-v4 DoubleDunkDeterministic-v4 DoubleDunkNoFrameskip-v4 DoubleDunk-ram-v0 DoubleDunk-ramDeterministic-v0 DoubleDunk-ramNoFrameskip-v0 DoubleDunk-ram-v4 DoubleDunk-ramDeterministic-v4 DoubleDunk-ramNoFrameskip-v4 ElevatorAction-v0 ElevatorActionDeterministic-v0 ElevatorActionNoFrameskip-v0 ElevatorAction-v4 ElevatorActionDeterministic-v4 ElevatorActionNoFrameskip-v4 ElevatorAction-ram-v0 ElevatorAction-ramDeterministic-v0 ElevatorAction-ramNoFrameskip-v0 ElevatorAction-ram-v4 ElevatorAction-ramDeterministic-v4 ElevatorAction-ramNoFrameskip-v4 Enduro-v0 EnduroDeterministic-v0 EnduroNoFrameskip-v0 Enduro-v4 EnduroDeterministic-v4 EnduroNoFrameskip-v4 Enduro-ram-v0 Enduro-ramDeterministic-v0 Enduro-ramNoFrameskip-v0 Enduro-ram-v4 Enduro-ramDeterministic-v4 Enduro-ramNoFrameskip-v4 FishingDerby-v0 FishingDerbyDeterministic-v0 FishingDerbyNoFrameskip-v0 FishingDerby-v4 FishingDerbyDeterministic-v4 FishingDerbyNoFrameskip-v4 FishingDerby-ram-v0 FishingDerby-ramDeterministic-v0 FishingDerby-ramNoFrameskip-v0 FishingDerby-ram-v4 FishingDerby-ramDeterministic-v4 FishingDerby-ramNoFrameskip-v4 Freeway-v0 FreewayDeterministic-v0 FreewayNoFrameskip-v0 Freeway-v4 FreewayDeterministic-v4 FreewayNoFrameskip-v4 Freeway-ram-v0 Freeway-ramDeterministic-v0 Freeway-ramNoFrameskip-v0 Freeway-ram-v4 Freeway-ramDeterministic-v4 Freeway-ramNoFrameskip-v4 Frostbite-v0 FrostbiteDeterministic-v0 FrostbiteNoFrameskip-v0 Frostbite-v4 FrostbiteDeterministic-v4 FrostbiteNoFrameskip-v4 Frostbite-ram-v0 Frostbite-ramDeterministic-v0 Frostbite-ramNoFrameskip-v0 Frostbite-ram-v4 Frostbite-ramDeterministic-v4 Frostbite-ramNoFrameskip-v4 Gopher-v0 GopherDeterministic-v0 GopherNoFrameskip-v0 Gopher-v4 GopherDeterministic-v4 GopherNoFrameskip-v4 Gopher-ram-v0 Gopher-ramDeterministic-v0 Gopher-ramNoFrameskip-v0 Gopher-ram-v4 Gopher-ramDeterministic-v4 Gopher-ramNoFrameskip-v4 Gravitar-v0 GravitarDeterministic-v0 GravitarNoFrameskip-v0 Gravitar-v4 GravitarDeterministic-v4 GravitarNoFrameskip-v4 Gravitar-ram-v0 Gravitar-ramDeterministic-v0 Gravitar-ramNoFrameskip-v0 Gravitar-ram-v4 Gravitar-ramDeterministic-v4 Gravitar-ramNoFrameskip-v4 Hero-v0 HeroDeterministic-v0 HeroNoFrameskip-v0 Hero-v4 HeroDeterministic-v4 HeroNoFrameskip-v4 Hero-ram-v0 Hero-ramDeterministic-v0 Hero-ramNoFrameskip-v0 Hero-ram-v4 Hero-ramDeterministic-v4 Hero-ramNoFrameskip-v4 IceHockey-v0 IceHockeyDeterministic-v0 IceHockeyNoFrameskip-v0 IceHockey-v4 IceHockeyDeterministic-v4 IceHockeyNoFrameskip-v4 IceHockey-ram-v0 IceHockey-ramDeterministic-v0 IceHockey-ramNoFrameskip-v0 IceHockey-ram-v4 IceHockey-ramDeterministic-v4 IceHockey-ramNoFrameskip-v4 Jamesbond-v0 JamesbondDeterministic-v0 JamesbondNoFrameskip-v0 Jamesbond-v4 JamesbondDeterministic-v4 JamesbondNoFrameskip-v4 Jamesbond-ram-v0 Jamesbond-ramDeterministic-v0 Jamesbond-ramNoFrameskip-v0 Jamesbond-ram-v4 Jamesbond-ramDeterministic-v4 Jamesbond-ramNoFrameskip-v4 JourneyEscape-v0 JourneyEscapeDeterministic-v0 JourneyEscapeNoFrameskip-v0 JourneyEscape-v4 JourneyEscapeDeterministic-v4 JourneyEscapeNoFrameskip-v4 JourneyEscape-ram-v0 JourneyEscape-ramDeterministic-v0 JourneyEscape-ramNoFrameskip-v0 JourneyEscape-ram-v4 JourneyEscape-ramDeterministic-v4 JourneyEscape-ramNoFrameskip-v4 Kangaroo-v0 KangarooDeterministic-v0 KangarooNoFrameskip-v0 Kangaroo-v4 KangarooDeterministic-v4 KangarooNoFrameskip-v4 Kangaroo-ram-v0 Kangaroo-ramDeterministic-v0 Kangaroo-ramNoFrameskip-v0 Kangaroo-ram-v4 Kangaroo-ramDeterministic-v4 Kangaroo-ramNoFrameskip-v4 Krull-v0 KrullDeterministic-v0 KrullNoFrameskip-v0 Krull-v4 KrullDeterministic-v4 KrullNoFrameskip-v4 Krull-ram-v0 Krull-ramDeterministic-v0 Krull-ramNoFrameskip-v0 Krull-ram-v4 Krull-ramDeterministic-v4 Krull-ramNoFrameskip-v4 KungFuMaster-v0 KungFuMasterDeterministic-v0 KungFuMasterNoFrameskip-v0 KungFuMaster-v4 KungFuMasterDeterministic-v4 KungFuMasterNoFrameskip-v4 KungFuMaster-ram-v0 KungFuMaster-ramDeterministic-v0 KungFuMaster-ramNoFrameskip-v0 KungFuMaster-ram-v4 KungFuMaster-ramDeterministic-v4 KungFuMaster-ramNoFrameskip-v4 MontezumaRevenge-v0 MontezumaRevengeDeterministic-v0 MontezumaRevengeNoFrameskip-v0 MontezumaRevenge-v4 MontezumaRevengeDeterministic-v4 MontezumaRevengeNoFrameskip-v4 MontezumaRevenge-ram-v0 MontezumaRevenge-ramDeterministic-v0 MontezumaRevenge-ramNoFrameskip-v0 MontezumaRevenge-ram-v4 MontezumaRevenge-ramDeterministic-v4 MontezumaRevenge-ramNoFrameskip-v4 MsPacman-v0 MsPacmanDeterministic-v0 MsPacmanNoFrameskip-v0 MsPacman-v4 MsPacmanDeterministic-v4 MsPacmanNoFrameskip-v4 MsPacman-ram-v0 MsPacman-ramDeterministic-v0 MsPacman-ramNoFrameskip-v0 MsPacman-ram-v4 MsPacman-ramDeterministic-v4 MsPacman-ramNoFrameskip-v4 NameThisGame-v0 NameThisGameDeterministic-v0 NameThisGameNoFrameskip-v0 NameThisGame-v4 NameThisGameDeterministic-v4 NameThisGameNoFrameskip-v4 NameThisGame-ram-v0 NameThisGame-ramDeterministic-v0 NameThisGame-ramNoFrameskip-v0 NameThisGame-ram-v4 NameThisGame-ramDeterministic-v4 NameThisGame-ramNoFrameskip-v4 Phoenix-v0 PhoenixDeterministic-v0 PhoenixNoFrameskip-v0 Phoenix-v4 PhoenixDeterministic-v4 PhoenixNoFrameskip-v4 Phoenix-ram-v0 Phoenix-ramDeterministic-v0 Phoenix-ramNoFrameskip-v0 Phoenix-ram-v4 Phoenix-ramDeterministic-v4 Phoenix-ramNoFrameskip-v4 Pitfall-v0 PitfallDeterministic-v0 PitfallNoFrameskip-v0 Pitfall-v4 PitfallDeterministic-v4 PitfallNoFrameskip-v4 Pitfall-ram-v0 Pitfall-ramDeterministic-v0 Pitfall-ramNoFrameskip-v0 Pitfall-ram-v4 Pitfall-ramDeterministic-v4 Pitfall-ramNoFrameskip-v4 Pong-v0 PongDeterministic-v0 PongNoFrameskip-v0 Pong-v4 PongDeterministic-v4 PongNoFrameskip-v4 Pong-ram-v0 Pong-ramDeterministic-v0 Pong-ramNoFrameskip-v0 Pong-ram-v4 Pong-ramDeterministic-v4 Pong-ramNoFrameskip-v4 Pooyan-v0 PooyanDeterministic-v0 PooyanNoFrameskip-v0 Pooyan-v4 PooyanDeterministic-v4 PooyanNoFrameskip-v4 Pooyan-ram-v0 Pooyan-ramDeterministic-v0 Pooyan-ramNoFrameskip-v0 Pooyan-ram-v4 Pooyan-ramDeterministic-v4 Pooyan-ramNoFrameskip-v4 PrivateEye-v0 PrivateEyeDeterministic-v0 PrivateEyeNoFrameskip-v0 PrivateEye-v4 PrivateEyeDeterministic-v4 PrivateEyeNoFrameskip-v4 PrivateEye-ram-v0 PrivateEye-ramDeterministic-v0 PrivateEye-ramNoFrameskip-v0 PrivateEye-ram-v4 PrivateEye-ramDeterministic-v4 PrivateEye-ramNoFrameskip-v4 Qbert-v0 QbertDeterministic-v0 QbertNoFrameskip-v0 Qbert-v4 QbertDeterministic-v4 QbertNoFrameskip-v4 Qbert-ram-v0 Qbert-ramDeterministic-v0 Qbert-ramNoFrameskip-v0 Qbert-ram-v4 Qbert-ramDeterministic-v4 Qbert-ramNoFrameskip-v4 Riverraid-v0 RiverraidDeterministic-v0 RiverraidNoFrameskip-v0 Riverraid-v4 RiverraidDeterministic-v4 RiverraidNoFrameskip-v4 Riverraid-ram-v0 Riverraid-ramDeterministic-v0 Riverraid-ramNoFrameskip-v0 Riverraid-ram-v4 Riverraid-ramDeterministic-v4 Riverraid-ramNoFrameskip-v4 RoadRunner-v0 RoadRunnerDeterministic-v0 RoadRunnerNoFrameskip-v0 RoadRunner-v4 RoadRunnerDeterministic-v4 RoadRunnerNoFrameskip-v4 RoadRunner-ram-v0 RoadRunner-ramDeterministic-v0 RoadRunner-ramNoFrameskip-v0 RoadRunner-ram-v4 RoadRunner-ramDeterministic-v4 RoadRunner-ramNoFrameskip-v4 Robotank-v0 RobotankDeterministic-v0 RobotankNoFrameskip-v0 Robotank-v4 RobotankDeterministic-v4 RobotankNoFrameskip-v4 Robotank-ram-v0 Robotank-ramDeterministic-v0 Robotank-ramNoFrameskip-v0 Robotank-ram-v4 Robotank-ramDeterministic-v4 Robotank-ramNoFrameskip-v4 Seaquest-v0 SeaquestDeterministic-v0 SeaquestNoFrameskip-v0 Seaquest-v4 SeaquestDeterministic-v4 SeaquestNoFrameskip-v4 Seaquest-ram-v0 Seaquest-ramDeterministic-v0 Seaquest-ramNoFrameskip-v0 Seaquest-ram-v4 Seaquest-ramDeterministic-v4 Seaquest-ramNoFrameskip-v4 Skiing-v0 SkiingDeterministic-v0 SkiingNoFrameskip-v0 Skiing-v4 SkiingDeterministic-v4 SkiingNoFrameskip-v4 Skiing-ram-v0 Skiing-ramDeterministic-v0 Skiing-ramNoFrameskip-v0 Skiing-ram-v4 Skiing-ramDeterministic-v4 Skiing-ramNoFrameskip-v4 Solaris-v0 SolarisDeterministic-v0 SolarisNoFrameskip-v0 Solaris-v4 SolarisDeterministic-v4 SolarisNoFrameskip-v4 Solaris-ram-v0 Solaris-ramDeterministic-v0 Solaris-ramNoFrameskip-v0 Solaris-ram-v4 Solaris-ramDeterministic-v4 Solaris-ramNoFrameskip-v4 SpaceInvaders-v0 SpaceInvadersDeterministic-v0 SpaceInvadersNoFrameskip-v0 SpaceInvaders-v4 SpaceInvadersDeterministic-v4 SpaceInvadersNoFrameskip-v4 SpaceInvaders-ram-v0 SpaceInvaders-ramDeterministic-v0 SpaceInvaders-ramNoFrameskip-v0 SpaceInvaders-ram-v4 SpaceInvaders-ramDeterministic-v4 SpaceInvaders-ramNoFrameskip-v4 StarGunner-v0 StarGunnerDeterministic-v0 StarGunnerNoFrameskip-v0 StarGunner-v4 StarGunnerDeterministic-v4 StarGunnerNoFrameskip-v4 StarGunner-ram-v0 StarGunner-ramDeterministic-v0 StarGunner-ramNoFrameskip-v0 StarGunner-ram-v4 StarGunner-ramDeterministic-v4 StarGunner-ramNoFrameskip-v4 Tennis-v0 TennisDeterministic-v0 TennisNoFrameskip-v0 Tennis-v4 TennisDeterministic-v4 TennisNoFrameskip-v4 Tennis-ram-v0 Tennis-ramDeterministic-v0 Tennis-ramNoFrameskip-v0 Tennis-ram-v4 Tennis-ramDeterministic-v4 Tennis-ramNoFrameskip-v4 TimePilot-v0 TimePilotDeterministic-v0 TimePilotNoFrameskip-v0 TimePilot-v4 TimePilotDeterministic-v4 TimePilotNoFrameskip-v4 TimePilot-ram-v0 TimePilot-ramDeterministic-v0 TimePilot-ramNoFrameskip-v0 TimePilot-ram-v4 TimePilot-ramDeterministic-v4 TimePilot-ramNoFrameskip-v4 Tutankham-v0 TutankhamDeterministic-v0 TutankhamNoFrameskip-v0 Tutankham-v4 TutankhamDeterministic-v4 TutankhamNoFrameskip-v4 Tutankham-ram-v0 Tutankham-ramDeterministic-v0 Tutankham-ramNoFrameskip-v0 Tutankham-ram-v4 Tutankham-ramDeterministic-v4 Tutankham-ramNoFrameskip-v4 UpNDown-v0 UpNDownDeterministic-v0 UpNDownNoFrameskip-v0 UpNDown-v4 UpNDownDeterministic-v4 UpNDownNoFrameskip-v4 UpNDown-ram-v0 UpNDown-ramDeterministic-v0 UpNDown-ramNoFrameskip-v0 UpNDown-ram-v4 UpNDown-ramDeterministic-v4 UpNDown-ramNoFrameskip-v4 Venture-v0 VentureDeterministic-v0 VentureNoFrameskip-v0 Venture-v4 VentureDeterministic-v4 VentureNoFrameskip-v4 Venture-ram-v0 Venture-ramDeterministic-v0 Venture-ramNoFrameskip-v0 Venture-ram-v4 Venture-ramDeterministic-v4 Venture-ramNoFrameskip-v4 VideoPinball-v0 VideoPinballDeterministic-v0 VideoPinballNoFrameskip-v0 VideoPinball-v4 VideoPinballDeterministic-v4 VideoPinballNoFrameskip-v4 VideoPinball-ram-v0 VideoPinball-ramDeterministic-v0 VideoPinball-ramNoFrameskip-v0 VideoPinball-ram-v4 VideoPinball-ramDeterministic-v4 VideoPinball-ramNoFrameskip-v4 WizardOfWor-v0 WizardOfWorDeterministic-v0 WizardOfWorNoFrameskip-v0 WizardOfWor-v4 WizardOfWorDeterministic-v4 WizardOfWorNoFrameskip-v4 WizardOfWor-ram-v0 WizardOfWor-ramDeterministic-v0 WizardOfWor-ramNoFrameskip-v0 WizardOfWor-ram-v4 WizardOfWor-ramDeterministic-v4 WizardOfWor-ramNoFrameskip-v4 YarsRevenge-v0 YarsRevengeDeterministic-v0 YarsRevengeNoFrameskip-v0 YarsRevenge-v4 YarsRevengeDeterministic-v4 YarsRevengeNoFrameskip-v4 YarsRevenge-ram-v0 YarsRevenge-ramDeterministic-v0 YarsRevenge-ramNoFrameskip-v0 YarsRevenge-ram-v4 YarsRevenge-ramDeterministic-v4 YarsRevenge-ramNoFrameskip-v4 Zaxxon-v0 ZaxxonDeterministic-v0 ZaxxonNoFrameskip-v0 Zaxxon-v4 ZaxxonDeterministic-v4 ZaxxonNoFrameskip-v4 Zaxxon-ram-v0 Zaxxon-ramDeterministic-v0 Zaxxon-ramNoFrameskip-v0 Zaxxon-ram-v4 Zaxxon-ramDeterministic-v4 Zaxxon-ramNoFrameskip-v4 CartPole-v0 CartPole-v1 MountainCar-v0 MountainCarContinuous-v0 Pendulum-v1 Acrobot-v1 LunarLander-v2 LunarLanderContinuous-v2 BipedalWalker-v3 BipedalWalkerHardcore-v3 CarRacing-v0 Blackjack-v1 FrozenLake-v1 FrozenLake8x8-v1 CliffWalking-v0 Taxi-v3 Reacher-v2 Pusher-v2 Thrower-v2 Striker-v2 InvertedPendulum-v2 InvertedDoublePendulum-v2 HalfCheetah-v2 HalfCheetah-v3 Hopper-v2 Hopper-v3 Swimmer-v2 Swimmer-v3 Walker2d-v2 Walker2d-v3 Ant-v2 Ant-v3 Humanoid-v2 Humanoid-v3 HumanoidStandup-v2 FetchSlide-v1 FetchPickAndPlace-v1 FetchReach-v1 FetchPush-v1 HandReach-v0 HandManipulateBlockRotateZ-v0 HandManipulateBlockRotateZTouchSensors-v0 HandManipulateBlockRotateZTouchSensors-v1 HandManipulateBlockRotateParallel-v0 HandManipulateBlockRotateParallelTouchSensors-v0 HandManipulateBlockRotateParallelTouchSensors-v1 HandManipulateBlockRotateXYZ-v0 HandManipulateBlockRotateXYZTouchSensors-v0 HandManipulateBlockRotateXYZTouchSensors-v1 HandManipulateBlockFull-v0 HandManipulateBlock-v0 HandManipulateBlockTouchSensors-v0 HandManipulateBlockTouchSensors-v1 HandManipulateEggRotate-v0 HandManipulateEggRotateTouchSensors-v0 HandManipulateEggRotateTouchSensors-v1 HandManipulateEggFull-v0 HandManipulateEgg-v0 HandManipulateEggTouchSensors-v0 HandManipulateEggTouchSensors-v1 HandManipulatePenRotate-v0 HandManipulatePenRotateTouchSensors-v0 HandManipulatePenRotateTouchSensors-v1 HandManipulatePenFull-v0 HandManipulatePen-v0 HandManipulatePenTouchSensors-v0 HandManipulatePenTouchSensors-v1 FetchSlideDense-v1 FetchPickAndPlaceDense-v1 FetchReachDense-v1 FetchPushDense-v1 HandReachDense-v0 HandManipulateBlockRotateZDense-v0 HandManipulateBlockRotateZTouchSensorsDense-v0 HandManipulateBlockRotateZTouchSensorsDense-v1 HandManipulateBlockRotateParallelDense-v0 HandManipulateBlockRotateParallelTouchSensorsDense-v0 HandManipulateBlockRotateParallelTouchSensorsDense-v1 HandManipulateBlockRotateXYZDense-v0 HandManipulateBlockRotateXYZTouchSensorsDense-v0 HandManipulateBlockRotateXYZTouchSensorsDense-v1 HandManipulateBlockFullDense-v0 HandManipulateBlockDense-v0 HandManipulateBlockTouchSensorsDense-v0 HandManipulateBlockTouchSensorsDense-v1 HandManipulateEggRotateDense-v0 HandManipulateEggRotateTouchSensorsDense-v0 HandManipulateEggRotateTouchSensorsDense-v1 HandManipulateEggFullDense-v0 HandManipulateEggDense-v0 HandManipulateEggTouchSensorsDense-v0 HandManipulateEggTouchSensorsDense-v1 HandManipulatePenRotateDense-v0 HandManipulatePenRotateTouchSensorsDense-v0 HandManipulatePenRotateTouchSensorsDense-v1 HandManipulatePenFullDense-v0 HandManipulatePenDense-v0 HandManipulatePenTouchSensorsDense-v0 HandManipulatePenTouchSensorsDense-v1 CubeCrash-v0 CubeCrashSparse-v0 CubeCrashScreenBecomesBlack-v0 MemorizeDigits-v0
1.2.2 安裝atari
pip install gym[atari]
pip uninstall atari_py
pip install --no-index -f https://github.com/Kojoley/atari-py/releases atari_py
1.2.3 安裝Gym Box2D
conda install -c anaconda swig
pip install box2d-py
1.2.4 安裝mujoco
1). Visual Studio安裝的時候要選擇windows 10 SDK
2). 在C:\Users\24410\下創建文件夾：.mujoco，在https://www.roboti.us/index.html下載mjpro150 win64，下載mjkey.txt，將https://www.apache.org/licenses/LICENSE-2.0.txt中的文字保存為LICENSE.txt。
3). 將這三個文件放在C:\Users\24410\.mujoco下。其中mjpro150 win64解壓，文件名為mjpro150。並將mjkey.txt與LICENSE.txt復制到C:\Users\24410\.mujoco\mjpro150\bin一份。
4). 添加系統環境變量
變量名：MUJOCO_PY_MJKEY_PATH
變量值：C:\Users\24410\.mujoco\mjpro150\bin\mjkey.txt
變量名：MUJOCO_PY_MUJOCO_PATH
變量值：C:\Users\24410\.mujoco\mjpro150\bin
並添加path路徑：C:\Users\24410\.mujoco\mjpro150\bin
5). 在終端輸入
cd C:\Users\24410\.mujoco\mjpro150\bin
simulate.exe ../model/humanoid.xml
6). 在RL環境下輸入pip install mujoco-py==1.50.1.68。完成。
2. 第一個Python小程序(gym環境)
2.1 gym環境的簡單使用(隨機采樣選擇動作)
這里給的是玩5局游戲，每局最多走1000步。
# -*- coding: UTF-8 -*-
# https://www.cnblogs.com/kailugaji/ - 凱魯嘎吉 - 博客園
import gym
import time
def run_gym(index):
env = gym.make(index)
for i_episode in range(5): # 玩幾局游戲
observation = env.reset() #用於重置環境
for t in range(1000): # # 每一局游戲最多1000步
env.render() # 用於渲染出當前的智能體以及環境的狀態
time.sleep(0.01) # 為了讓顯示變慢，否則畫面會非常快
action = env.action_space.sample() # 隨機采樣選擇動作，這一步后續可以通過RL策略獲得好的動作，而不是隨機
observation, reward, done, info = env.step(action) # take a random action
"""
env.step()返回四個值：
observation（object）一個特定的環境對象，代表了你從環境中得到的觀測值
例如從攝像頭獲得的像素數據，機器人的關節角度和關節速度，或者棋盤游戲的棋盤
reward（float）由於之前采取的動作所獲得的大量獎勵，與環境交互的過程中，獎勵值的規模會發生變化，但是總體的目標一直都是使得總獎勵最大
done（boolean）決定是否將環境初始化，大多數，但不是所有的任務都被定義好了什么情況該結束這個回合
例如倒立擺的小車離開地太遠了就結束了這個回合
info（dict）調試過程中將會產生的有用信息，有時它會對我們的強化學習學習過程很有用
例如，有時它會包含最后一個狀態改變后的原始概率
然而在評估你的智能體的時候你是不會用到這些信息去驅動你的智能體學習的
"""
if done:
print("Episode finished after {} timesteps".format(t+1))
break
env.close()
return observation, reward, env.action_space, env.observation_space
if __name__ == "__main__":
index = 'CartPole-v0' # Classic control
# index = 'MountainCar-v0' # Classic control
# index = 'AirRaid-ram-v0' # Atari
# index = 'Taxi-v3' # Toy text
'''
出租車調度
這里有 4 個地點，分別用 4 個字母表示，任務是要從一個地點接上乘客，送到另外 3 個中的一個放下乘客，越快越好。
顏色：藍色：乘客，紅色：乘客的目的地，黃色：空出租車，綠色：出租車滿座，其中 “:” 柵欄可以穿越，"|" 柵欄不能穿越
Reward: 成功運送一個客人獲得 20 分獎勵
每走一步損失 1 分（希望盡快送到目的地）
沒有把客人放到指定的位置，損失 10 分
Action: 0：向南移動，1：向北移動，2：向東移動，3：向西移動，4：乘客上車，5：乘客下車
State: 500維，（出租車行、出租車列、乘客位置、目的地）
'''
# index = 'Ant-v2'
# index = 'BipedalWalker-v3' # Box2D
'''
訓練兩足機器人行走
Goal：Agent需要學會克服各種障礙向前移動
State: 24維向量，包括各部件角速度，水平速度，垂直速度，關節位置，腿與地面的接觸以及10個激光雷達測距儀的測量值
Action: 4維連續動作空間，取值范圍為[-1,1]，分別對應機器人胯下兩個關節的轉矩以及膝關節的轉矩
Reward: 向前移動會獲得到正獎勵信號，摔倒會得到-100的獎勵信號，同時，驅動各關節轉動會得到一定的負獎勵信號
Done: 摔倒或抵達地圖終點會結束當前回合
'''
# index = 'LunarLander-v2' # Box2D
'''
將着陸器導航到其着陸台
着陸點始終位於坐標 (0,0)。坐標是狀態向量中的前兩個數字。
燃料是無限的，所以代理可以學習飛行，然后在第一次嘗試時着陸。
Reward: 從屏幕頂部移動到着陸墊和零速度的獎勵約為 100..140 點。
如果着陸器遠離着陸台，它會失去獎勵。如果着陸器墜毀或靜止，情節結束，獲得額外的 -100 或 +100 分。
每條腿接地是+10。點火主機每幀-0.3分。解決是200分。可以在着陸場外着陸。
Action: 什么都不做，向左方向引擎開火，向主引擎開火，向右方向引擎開火。
State: 水平坐標x，垂直坐標y，水平速度，垂直速度，角度，角速度，腿1觸地，腿2觸地。
'''
observation, reward, action_space, observation_space = run_gym(index)
print('Action Space: \n', action_space)
print('Observation Space: \n', observation_space)
print('Observation: \n', observation)
print('Reward: \n', reward)
CartPole-v0
MountainCar-v0
AirRaid-ram-v0
Taxi-v3
Ant-v2
BipedalWalker-v3
LunarLander-v2
2.2 將gym游戲界面保存為gif動圖
這里給的是玩1局游戲，每局最多走1000步。
# -*- coding: UTF-8 -*-
# https://www.cnblogs.com/kailugaji/ - 凱魯嘎吉 - 博客園
import gym
import time
from matplotlib import animation
import matplotlib.pyplot as plt
# 將gym界面保存為gif動圖
def save_frames_as_gif(frames, path, index):
filename = 'gym_'+ index + '.gif'
#Mess with this to change frame size
plt.figure(figsize=(frames[0].shape[1] / 72.0, frames[0].shape[0] / 72.0), dpi=72)
patch = plt.imshow(frames[0])
plt.axis('off')
def animate(i):
patch.set_data(frames[i])
anim = animation.FuncAnimation(plt.gcf(), animate, frames = len(frames), interval=50)
anim.save(path + filename, writer='pillow', fps=60)
def run_gym(index):
env = gym.make(index)
frames = []
for i_episode in range(1): # 玩幾局游戲
observation = env.reset() #用於重置環境
for t in range(1000): # # 每一局游戲最多1000步
env.render() # 用於渲染出當前的智能體以及環境的狀態
frames.append(env.render(mode='rgb_array'))
time.sleep(0.01) # 為了讓顯示變慢，否則畫面會非常快
action = env.action_space.sample() # 隨機采樣選擇動作，這一步后續可以通過RL策略獲得好的動作，而不是隨機
observation, reward, done, info = env.step(action) # take a random action
if done:
print("Episode finished after {} timesteps".format(t+1))
break
env.close()
save_frames_as_gif(frames, path = './', index = index)
return observation, reward, env.action_space, env.observation_space
if __name__ == "__main__":
index = 'CartPole-v0' # Classic control
# index = 'MountainCar-v0' # Classic control
# index = 'AirRaid-ram-v0' # Atari
# index = 'Taxi-v3' # Toy text
# index = 'Ant-v2'
# index = 'BipedalWalker-v3' # Box2D
# index = 'LunarLander-v2' # Box2D
observation, reward, action_space, observation_space = run_gym(index)
print('Action Space: \n', action_space)
print('Observation Space: \n', observation_space)
print('Observation: \n', observation)
print('Reward: \n', reward)
CartPole-v0
Episode finished after 14 timesteps
Action Space:
Discrete(2)
Observation Space:
Box([-4.8000002e+00 -3.4028235e+38 -4.1887903e-01 -3.4028235e+38], [4.8000002e+00 3.4028235e+38 4.1887903e-01 3.4028235e+38], (4,), float32)
Observation:
[-0.21281277 -0.82338583 0.21441801 1.4051291 ]
Reward:
1.0
MountainCar-v0
Episode finished after 200 timesteps
Action Space:
Discrete(3)
Observation Space:
Box([-1.2 -0.07], [0.6 0.07], (2,), float32)
Observation:
[-0.44918552 0.00768781]
Reward:
-1.0
AirRaid-ram-v0
Episode finished after 520 timesteps
Action Space:
Discrete(6)
Observation Space:
Box([0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0], [255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255
255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255
255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255
255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255
255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255
255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255
255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255 255
255 255], (128,), uint8)
Observation:
[188 184 180 184 188 62 188 0 0 0 128 16 56 68 146 56 68 0
37 64 44 124 0 191 252 175 252 2 1 240 240 240 62 62 62 0
0 0 224 255 0 8 117 14 14 4 155 246 150 246 145 246 140 246
135 246 130 246 234 0 31 31 10 1 2 6 30 79 123 0 0 0
0 2 2 0 16 16 181 236 236 0 0 248 248 35 89 1 0 3
0 0 22 60 80 80 64 0 2 15 0 0 0 0 0 0 0 0
0 0 0 0 62 247 236 247 62 247 181 247 240 240 0 0 202 245
144 245]
Reward:
0.0
Ant-v2
Episode finished after 84 timesteps
Action Space:
Box([-1. -1. -1. -1. -1. -1. -1. -1.], [1. 1. 1. 1. 1. 1. 1. 1.], (8,), float32)
Observation Space:
Box([-inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf
-inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf
-inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf
-inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf
-inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf
-inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf
-inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf
-inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf], [inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf
inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf
inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf
inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf
inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf
inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf
inf inf inf], (111,), float64)
Observation:
[ 1.01382928 0.93552107 -0.3281438 -0.11665695 0.05927168 0.45597194
0.97790293 0.54600325 -1.25270942 -0.52486369 -1.22324054 -0.37710358
1.20411015 -0.63179069 0.77311549 1.25462637 -3.07764597 -0.53052459
-0.56717237 5.52724079 -7.48125022 -0.67642035 1.00847733 0.0252006
-6.03428845 5.96745216 5.86833084 0. 0. 0.
0. 0. 0. 0. 0. 0.
0. 0. 0. 0. 0. 0.
0. 0. 0. 0. 0. 0.
0. 0. 0. 0. 0. 0.
0. 0. 0. 0. 0. 0.
0. 0. 0. 0. 0. 0.
0. 0. 0. 0. 0. 0.
0. 0. 0. 0. 0. 0.
0. 0. 0. 0. 0. 0.
0. 0. 0. 0. 0. 0.
0. 0. 0. 0. 0. 0.
0. 0. 0. 0. 0. 0.
0. 0. 0. 0. 0. 0.
0. 0. 0. ]
Reward:
-1.3955676167834756
BipedalWalker-v3
Episode finished after 72 timesteps
Action Space:
Box([-1. -1. -1. -1.], [1. 1. 1. 1.], (4,), float32)
Observation Space:
Box([-inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf -inf
-inf -inf -inf -inf -inf -inf -inf -inf -inf -inf], [inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf inf
inf inf inf inf inf inf], (24,), float32)
Observation:
[ 2.2528927 0.10642751 -0.30818883 -0.07858526 -0.64701635 0.07910275
-0.09122896 -0.10672931 1. -0.7459374 -1.264112 -0.5282707
0.41768146 1. 0.18380441 0.18589178 0.19239755 0.20412572
0.22270261 0.25120568 0.2956909 0.36940333 0.50724566 0.83926386]
Reward:
-100
LunarLander-v2
Episode finished after 88 timesteps
Action Space:
Discrete(4)
Observation Space:
Box([-inf -inf -inf -inf -inf -inf -inf -inf], [inf inf inf inf inf inf inf inf], (8,), float32)
Observation:
[ 0.620516 -0.04573616 0.10586149 0.08483806 -1.4809935 -0.49739528
0. 0. ]
Reward:
-100
由於這里是隨機選動作，所以基本沒有學習到任何東西，全靠蒙。后續可以通過強化學習的相關方法來確定agent的下一步動作。
atari, mujoco在第一個程序里渲染的顏色沒有問題，但第二個程序里渲染的顏色出問題了，但不影響算法的學習能力(雖然這里並未學習)，可能是某些包版本的問題，有待進一步改善。
推薦網址(介紹gym部分環境)：https://www.gymlibrary.dev/