Modular Multi-Objective Deep Reinforcement Learning with Decision Values

被引:0
|
作者
Tajmajer, Tomasz [1 ]
机构
[1] Univ Warsaw, Inst Informat, Ul Banacha 2, PL-02097 Warsaw, Poland
关键词
D O I
暂无
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
In this work we present a method for using Deep Q-Networks (DQNs) in multi-objective environments. Deep Q-Networks provide remarkable performance in single objective problems learning from high-level visual state representations. However, in many scenarios (e.g in robotics, games), the agent needs to pursue multiple objectives simultaneously. We propose an architecture in which separate DQNs are used to control the agent's behaviour with respect to particular objectives. In this architecture we introduce decision values to improve the scalarization of multiple DQNs into a single action. Our architecture enables the decomposition of the agent's behaviour into controllable and replaceable sub-behaviours learned by distinct modules. Moreover, it allows to change the priorities of particular objectives post-learning, while preserving the overall performance of the agent. To evaluate our solution we used a game-like simulator in which an agent - provided with high-level visual input - pursues multiple objectives in a 2D world.
引用
收藏
页码:85 / 93
页数:9
相关论文
共 50 条
  • [1] A multi-objective deep reinforcement learning framework
    Thanh Thi Nguyen
    Ngoc Duy Nguyen
    Vamplew, Peter
    Nahavandi, Saeid
    Dazeley, Richard
    Lim, Chee Peng
    [J]. ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE, 2020, 96
  • [2] Dynamic Weights in Multi-Objective Deep Reinforcement Learning
    Abels, Axel
    Roijers, Diederik M.
    Lenaerts, Tom
    Nowe, Ann
    Steckelmacher, Denis
    [J]. INTERNATIONAL CONFERENCE ON MACHINE LEARNING, VOL 97, 2019, 97
  • [3] Urban Driving with Multi-Objective Deep Reinforcement Learning
    Li, Changjian
    Czarnecki, Krzysztof
    [J]. AAMAS '19: PROCEEDINGS OF THE 18TH INTERNATIONAL CONFERENCE ON AUTONOMOUS AGENTS AND MULTIAGENT SYSTEMS, 2019, : 359 - 367
  • [4] Multi-objective safe reinforcement learning: the relationship between multi-objective reinforcement learning and safe reinforcement learning
    Horie, Naoto
    Matsui, Tohgoroh
    Moriyama, Koichi
    Mutoh, Atsuko
    Inuzuka, Nobuhiro
    [J]. ARTIFICIAL LIFE AND ROBOTICS, 2019, 24 (03) : 352 - 359
  • [5] Multi-objective safe reinforcement learning: the relationship between multi-objective reinforcement learning and safe reinforcement learning
    Naoto Horie
    Tohgoroh Matsui
    Koichi Moriyama
    Atsuko Mutoh
    Nobuhiro Inuzuka
    [J]. Artificial Life and Robotics, 2019, 24 : 352 - 359
  • [6] Multi-objective path planning based on deep reinforcement learning
    Xu, Jian
    Huang, Fei
    Cui, Yunfei
    Du, Xue
    [J]. 2022 41ST CHINESE CONTROL CONFERENCE (CCC), 2022, : 3273 - 3279
  • [7] Deep reinforcement learning for multi-objective game strategy selection
    Jiang, Ruhao
    Deng, Yanchen
    Chen, Yingying
    Luo, He
    An, Bo
    [J]. COMPUTERS & OPERATIONS RESEARCH, 2024, 168
  • [8] Multi-condition multi-objective optimization using deep reinforcement learning
    Kim, Sejin
    Kim, Innyoung
    You, Donghyun
    [J]. JOURNAL OF COMPUTATIONAL PHYSICS, 2022, 462
  • [9] Multi-objective vehicle following decision algorithm based on reinforcement learning
    Dend, Xiao-Hao
    Hou, Jin
    Tan, Guang-Hong
    Wan, Bin-Yang
    Cao, Ting-Ting
    [J]. Kongzhi yu Juece/Control and Decision, 2021, 36 (10): : 2497 - 2503
  • [10] Multi-objective ω-Regular Reinforcement Learning
    Hahn, Ernst Moritz
    Perez, Mateo
    Schewe, Sven
    Somenzi, Fabio
    Trivedi, Ashutosh
    Wojtczak, Dominik
    [J]. FORMAL ASPECTS OF COMPUTING, 2023, 35 (02)