FlowPG: Action-constrained policy gradient with normalizing flows

• المؤلفون الرئيسيون: BRAHMANAGE JANAKA CHATHURANGA THILAKARATHNA, LING, Jiajing, KUMAR, Akshat
• التنسيق: text
• اللغة: English
• منشور في: Institutional Knowledge at Singapore Management University 2023
• الموضوعات: Artificial Intelligence and Robotics; Databases and Information Systems
• الوصول للمادة أونلاين: https://ink.library.smu.edu.sg/sis_research/8551; https://ink.library.smu.edu.sg/context/sis_research/article/9554/viewcontent/11351_flowpg_action_constrained_poli.pdf

Action-constrained reinforcement learning (ACRL) is a popular approach for solving safety-critical and resource-allocation related decision making problems. A major challenge in ACRL is to ensure agent taking a valid action satisfying constraints in each RL step. Commonly used approach of using a pr...