Adversarial Robust Deep Reinforcement Learning Requires Redefining Robustness

被引:0
|
作者
Korkmaz, Ezgi
机构
关键词
LEVEL;
D O I
暂无
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Learning from raw high dimensional data via interaction with a given environment has been effectively achieved through the utilization of deep neural networks. Yet the observed degradation in policy performance caused by imperceptible worst-case policy dependent translations along high sensitivity directions (i.e. adversarial perturbations) raises concerns on the robustness of deep reinforcement learning policies. In our paper, we show that these high sensitivity directions do not lie only along particular worst-case directions, but rather are more abundant in the deep neural policy landscape and can be found via more natural means in a black-box setting. Furthermore, we show that vanilla training techniques intriguingly result in learning more robust policies compared to the policies learnt via the state-of-the-art adversarial training techniques. We believe our work lays out intriguing properties of the deep reinforcement learning policy manifold and our results can help to build robust and generalizable deep reinforcement learning policies.
引用
收藏
页码:8369 / 8377
页数:9
相关论文
共 50 条
  • [41] Improved Robustness and Safety for Autonomous Vehicle Control with Adversarial Reinforcement Learning
    Ma, Xiaobai
    Driggs-Campbell, Katherine
    Kochenderfer, Mykel J.
    2018 IEEE INTELLIGENT VEHICLES SYMPOSIUM (IV), 2018, : 1665 - 1671
  • [42] Trade-Off Between Robustness and Rewards Adversarial Training for Deep Reinforcement Learning Under Large Perturbations
    Huang, Jeffrey
    Choi, Ho Jin
    Figueroa, Nadia
    IEEE ROBOTICS AND AUTOMATION LETTERS, 2023, 8 (12) : 8018 - 8025
  • [43] Understanding adversarial attacks on observations in deep reinforcement learning
    You, Qiaoben
    Ying, Chengyang
    Zhou, Xinning
    Su, Hang
    Zhu, Jun
    Zhang, Bo
    SCIENCE CHINA-INFORMATION SCIENCES, 2024, 67 (05)
  • [44] Deep Adversarial Reinforcement Learning With Noise Compensation by Autoencoder
    Ohashi, Kohei
    Nakanishi, Kosuke
    Sasaki, Wataru
    Yasui, Yuji
    Ishii, Shin
    IEEE ACCESS, 2021, 9 : 143901 - 143912
  • [45] Tactics of Adversarial Attack on Deep Reinforcement Learning Agents
    Lin, Yen-Chen
    Hong, Zhang-Wei
    Liao, Yuan-Hong
    Shih, Meng-Li
    Liu, Ming-Yu
    Sun, Min
    PROCEEDINGS OF THE TWENTY-SIXTH INTERNATIONAL JOINT CONFERENCE ON ARTIFICIAL INTELLIGENCE, 2017, : 3756 - 3762
  • [46] Attacking Deep Reinforcement Learning With Decoupled Adversarial Policy
    Mo, Kanghua
    Tang, Weixuan
    Li, Jin
    Yuan, Xu
    IEEE TRANSACTIONS ON DEPENDABLE AND SECURE COMPUTING, 2023, 20 (01) : 758 - 768
  • [47] A Survey on Adversarial Attacks and Defenses for Deep Reinforcement Learning
    Liu A.-S.
    Guo J.
    Li S.-M.
    Xiao Y.-S.
    Liu X.-L.
    Tao D.-C.
    Jisuanji Xuebao/Chinese Journal of Computers, 2023, 46 (08): : 1553 - 1576
  • [48] On the Robustness of Controlled Deep Reinforcement Learning for Slice Placement
    Jose Jurandir Alves Esteves
    Amina Boubendir
    Fabrice Guillemin
    Pierre Sens
    Journal of Network and Systems Management, 2022, 30
  • [49] Understanding adversarial attacks on observations in deep reinforcement learning
    You QIAOBEN
    Chengyang YING
    Xinning ZHOU
    Hang SU
    Jun ZHU
    Bo ZHANG
    Science China(Information Sciences), 2024, 67 (05) : 69 - 83
  • [50] Challenges and Countermeasures for Adversarial Attacks on Deep Reinforcement Learning
    Ilahi I.
    Usama M.
    Qadir J.
    Janjua M.U.
    Al-Fuqaha A.
    Hoang D.T.
    Niyato D.
    IEEE Transactions on Artificial Intelligence, 2022, 3 (02): : 90 - 109