Deep Predictive Policy Training using Reinforcement Learning

被引:0
|
作者
Ghadirzadeh, Ali [1 ]
Maki, Atsuto [1 ]
Kragic, Danica [1 ]
Bjorkman, Marten [1 ]
机构
[1] KTH Royal Inst Technol, CSC, Robot Percept & Learning Lab RPL, Stockholm, Sweden
来源
2017 IEEE/RSJ INTERNATIONAL CONFERENCE ON INTELLIGENT ROBOTS AND SYSTEMS (IROS) | 2017年
基金
瑞典研究理事会; 欧盟地平线“2020”;
关键词
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
Skilled robot task learning is best implemented by predictive action policies due to the inherent latency of sensorimotor processes. However, training such predictive policies is challenging as it involves finding a trajectory of motor activations for the full duration of the action. We propose a data-efficient deep predictive policy training (DPPT) framework with a deep neural network policy architecture which maps an image observation to a sequence of motor activations. The architecture consists of three sub-networks referred to as the perception, policy and behavior super-layers. The perception and behavior super-layers force an abstraction of visual and motor data trained with synthetic and simulated training samples, respectively. The policy super-layer is a small sub-network with fewer parameters that maps data in-between the abstracted manifolds. It is trained for each task using methods for policy search reinforcement learning. We demonstrate the suitability of the proposed architecture and learning framework by training predictive policies for skilled object grasping and ball throwing on a PR2 robot. The effectiveness of the method is illustrated by the fact that these tasks are trained using only about 180 real robot attempts with qualitative terminal rewards.
引用
收藏
页码:2351 / 2358
页数:8
相关论文
共 50 条
  • [1] Adversarial Policy Training against Deep Reinforcement Learning
    Wu, Xian
    Guo, Wenbo
    Wei, Hua
    Xing, Xinyu
    PROCEEDINGS OF THE 30TH USENIX SECURITY SYMPOSIUM, 2021, : 1883 - 1900
  • [2] ROBOTIC GRASPING TRAINING USING DEEP REINFORCEMENT LEARNING WITH POLICY GUIDANCE MECHANISM
    Yao, Junying
    Liu, Yongkui
    Lin, Tingyu
    Ping, Xubin
    Xu, He
    Wang, Wenxiao
    Xiao, Yingying
    Zhang, Lin
    Wang, Lihui
    PROCEEDINGS OF THE ASME 2021 16TH INTERNATIONAL MANUFACTURING SCIENCE AND ENGINEERING CONFERENCE (MSEC2021), VOL 2, 2021,
  • [3] On Training Flexible Robots using Deep Reinforcement Learning
    Dwiel, Zach
    Candadai, Madhavun
    Phielipp, Mariano
    2019 IEEE/RSJ INTERNATIONAL CONFERENCE ON INTELLIGENT ROBOTS AND SYSTEMS (IROS), 2019, : 4666 - 4671
  • [4] Policy Reuse in Deep Reinforcement Learning
    Glatt, Ruben
    Helena, Anna
    Costa, Reali
    THIRTY-FIRST AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE, 2017, : 4929 - 4930
  • [5] Reaching Pruning Locations in a Vine Using a Deep Reinforcement Learning Policy
    Yandun, Francisco
    Parhar, Tanvir
    Silwal, Abhisesh
    Clifford, David
    Yuan, Zhiqiang
    Levine, Gabriella
    Yaroshenko, Sergey
    Kantor, George
    2021 IEEE INTERNATIONAL CONFERENCE ON ROBOTICS AND AUTOMATION (ICRA 2021), 2021, : 2400 - 2406
  • [6] Probabilistic Policy Blending for Shared Autonomy using Deep Reinforcement Learning
    Singh, Saurav
    Heard, Jamison
    2023 32ND IEEE INTERNATIONAL CONFERENCE ON ROBOT AND HUMAN INTERACTIVE COMMUNICATION, RO-MAN, 2023, : 1537 - 1544
  • [7] Training and Evaluation of Deep Policies Using Reinforcement Learning and Generative Models
    Ghadirzadeh, Ali
    Poklukar, Petra
    Arndt, Karol
    Finn, Chelsea
    Kyrki, Ville
    Kragic, Danica
    Björkman, Mårten
    Journal of Machine Learning Research, 2022, 23
  • [8] Robot-Assisted Training in Laparoscopy Using Deep Reinforcement Learning
    Tan, Xiaoyu
    Chng, Chin-Boon
    Su, Ye
    Lim, Kah-Bin
    Chui, Chee-Kong
    IEEE ROBOTICS AND AUTOMATION LETTERS, 2019, 4 (02) : 485 - 492
  • [9] Training and Evaluation of Deep Policies Using Reinforcement Learning and Generative Models
    Ghadirzadeh, Ali
    Poklukar, Petra
    Arndt, Karol
    Finn, Chelsea
    Kyrki, Ville
    Kragic, Danica
    Bjorkman, Marten
    JOURNAL OF MACHINE LEARNING RESEARCH, 2022, 23
  • [10] Diversity Evolutionary Policy Deep Reinforcement Learning
    Liu, Jian
    Feng, Liming
    COMPUTATIONAL INTELLIGENCE AND NEUROSCIENCE, 2021, 2021