Hollow_Knight_RL

Link

https://www.youtube.com/watch?v=EeooyhH2kHc

修正

• 2024/05/22 修正模型
• 2024/05/23 模型探索測試(200次，勝率0%)
• 2024/05/24 模型測試(200次，勝率6%)
• 2024/05/25 模型修正(200次，勝率50%)
• 2024/05/26 修正環境
• 2024/05/29 發現獎勵有問題，修正擷取玩家血量的方式

Action

攻擊

動作名稱	動作函數	按鍵
攻擊	Attack()	X
向上攻擊	Attack_Up()	UP + X
短跳躍	Short_Jump()	C + DOWN
中跳躍	Mid_Jump()	C
向上技能	Skill_Up()	UP + Z
向下技能	Skill_Down()	DOWN + Z
衝刺	Rush()	左 Shift
治療	Cure()	A

移動

動作名稱	動作函數	按鍵
沒有任何動作	Nothing()	無
向左	Move_Left()	LEFT(long)
向右	Move_Right()	RIGHT(long)
轉頭左	Turn_Left()	LEFT(short)
轉頭右	Turn_Right()	RIGHT(short)

References

• Playing Atari with Deep Reinforcement Learning

• Deep Reinforcement Learning with Double Q-learning

• Dueling Network Architectures for Deep Reinforcement Learning

• Spectral Normalisation for Deep Reinforcement Learning: an Optimisation Perspective

• Rainbow: Combining Improvements in Deep Reinforcement Learning

• Image Augmentation Is All You Need: Regularizing Deep Reinforcement Learning from Pixels

• Deep Residual Learning for Image Recognition

• Squeeze-and-Excitation Networks

• A ConvNet for the 2020s

• Bag of Tricks for Image Classification with Convolutional Neural Networks

• Noisy Networks for Exploration

• The Primacy Bias in Deep Reinforcement Learning

• Prioritized Experience Replay

• Stabilizing Deep Q-Learning with ConvNets and Vision Transformers under Data Augmentation

• Averaged-DQN: Variance Reduction and Stabilization for Deep Reinforcement Learning

• https://github.com/toshikwa/fqf-iqn-qrdqn.pytorch/blob/master/fqf_iqn_qrdqn/network.py

• https://github.com/pytorch/vision/blob/main/torchvision/models/vgg.py

• https://www.youtube.com/watch?v=NP8pXZdU-5U

• https://github.com/seermer/HollowKnight_RL

• https://github.com/ailec0623/DQN_HollowKnight