Safe Reinforcement Learning using Data-Driven Predictive Control

被引:1
|
作者
Selim, Mahmoud [1 ]
Alanwar, Amr [2 ]
El-Kharashi, M. Watheq [1 ]
Abbas, Hazem M. [1 ]
Johansson, Karl H. [3 ]
机构
[1] Ain Shams Univ, Cairo, Egypt
[2] Jacobs Univ, Bremen, Germany
[3] KTH Royal Inst Technol, Stockholm, Sweden
关键词
Reinforcement learning; robot safety; task and motion planning;
D O I
10.1109/ICCSPA55860.2022.10018994
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Reinforcement learning (RL) algorithms can achieve state-of-the-art performance in decision-making and continuous control tasks. However, applying RL algorithms on safety-critical systems still needs to be well justified due to the exploration nature of many RL algorithms, especially when the model of the robot and the environment are unknown. To address this challenge, we propose a data-driven safety layer that acts as a filter for unsafe actions. The safety layer uses a data-driven predictive controller to enforce safety guarantees for RL policies during training and after deployment. The RL agent proposes an action that is verified by computing the data-driven reachability analysis. If there is an intersection between the reachable set of the robot using the proposed action, we call the data-driven predictive controller to find the closest safe action to the proposed unsafe action. The safety layer penalizes the RL agent if the proposed action is unsafe and replaces it with the closest safe one. In the simulation, we show that our method outperforms state-of-the-art safe RL methods on the robotics navigation problem for a Turtlebot 3 in Gazebo and a quadrotor in Unreal Engine 4 (UE4).
引用
收藏
页数:6
相关论文
共 50 条
  • [21] A survey on load frequency control using reinforcement learning-based data-driven controller
    Muduli, Rasananda
    Jena, Debashisha
    Moger, Tukaram
    [J]. APPLIED SOFT COMPUTING, 2024, 166
  • [22] Sparse Wide-Area Control of Power Systems using Data-driven Reinforcement Learning
    Dizche, Amirhassan Fallah
    Chakrabortty, Aranya
    Duel-Hallen, Alexandra
    [J]. 2019 AMERICAN CONTROL CONFERENCE (ACC), 2019, : 2867 - 2872
  • [23] Robust data-driven predictive control using reachability analysis
    Alanwar, Amr
    Stuerz, Yvonne
    Johansson, Karl Henrik
    [J]. EUROPEAN JOURNAL OF CONTROL, 2022, 68
  • [24] Data-driven predictive point-to-point iterative learning control
    Zhang, Xueming
    Hou, Zhongsheng
    [J]. NEUROCOMPUTING, 2023, 518 : 431 - 439
  • [25] Data-driven Modelling, Learning and Stochastic Predictive Control for the Steel Industry
    Herceg, Domagoj
    Georgoulas, George
    Sopasakis, Pantelis
    Castano, Miguel
    Patrinos, Panagiotis
    Bemporad, Alberto
    Niemi, Jan
    Nikolakopoulos, George
    [J]. 2017 25TH MEDITERRANEAN CONFERENCE ON CONTROL AND AUTOMATION (MED), 2017, : 1361 - 1366
  • [26] Data-driven Iterative Learning for Model Predictive Control of Heating Systems
    Lautenschlager, Bjoern
    Lichtenberg, Gerwald
    [J]. IFAC PAPERSONLINE, 2016, 49 (13): : 175 - 180
  • [27] Data-Driven Passivity Analysis and Fault Detection Using Reinforcement Learning
    Ma, Haoran
    Zhao, Zhengen
    Li, Zhuyuan
    Yang, Ying
    [J]. IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS I-REGULAR PAPERS, 2024,
  • [28] Data-Driven Design of a Reference Governor Using Deep Reinforcement Learning
    Angelica Taylor, Maria
    Felipe Giraldo, Luis
    [J]. 5TH IEEE CONFERENCE ON CONTROL TECHNOLOGY AND APPLICATIONS (IEEE CCTA 2021), 2021, : 956 - 961
  • [29] Data-driven constrained reinforcement learning algorithm for path tracking control of hovercraft
    Wang, Yuanhui
    Zhou, Hua
    [J]. OCEAN ENGINEERING, 2024, 307
  • [30] Underactuated MIMO Airship Control Based on Online Data-Driven Reinforcement Learning
    Boase, Derek
    Gueaieb, Wail
    Miah, Md Suruz
    [J]. 2023 IEEE/RSJ INTERNATIONAL CONFERENCE ON INTELLIGENT ROBOTS AND SYSTEMS (IROS), 2023, : 9464 - 9471