Action Space Reinforcement Learning

Action space reinforcement learning (RL) focuses on designing and optimizing the set of actions available to an RL agent to improve learning efficiency and performance in various tasks. Current research emphasizes adapting action spaces to specific problem structures, such as incorporating parallel actions or dynamically selecting actions from expanding sets, often employing algorithms like Q-learning, TD3, and policy gradient methods. This approach is proving valuable in diverse applications, including robotics (e.g., drone control and inverted pendulum balancing), natural language processing (topic modeling), and finance (algorithmic trading), demonstrating the power of tailored action spaces for solving complex real-world problems.

Papers

March 13, 2024

Meta-operators for Enabling Parallel Planning Using Deep Reinforcement Learning
Ángel Aso-Mollar, Eva Onaindia
Reinforcement Learning Deep Reinforcement Learning Markov Decision Process Task Planning Generalized Planning Action Space Reinforcement Learning

December 5, 2023

A Q-learning approach to the continuous control problem of robot inverted pendulum balancing
Mohammad Safeea, Pedro Neto
Robot Person Q Learning Action Space Control Problem Inverted Pendulum Action Space Reinforcement Learning

May 8, 2023

Reinforcement Learning for Topic Models
Jeremy Costello, Marek Z. Reformat
Reinforcement Learning Contrastive Learning Variational Autoencoder Topic Model Action Space Reinforcement Learning

April 3, 2023

Action Pick-up in Dynamic Action Space Reinforcement Learning
Jiaqi Ye, Xiaodong Li, Pangjing Wu, Feng Wang
Markov Decision Process Action Space Action Selection Multi Commodity Pickup Action Space Reinforcement Learning

October 7, 2022

Algorithmic Trading Using Continuous Action Space Deep Reinforcement Learning
Naseh Majidi, Mahdi Shamsi, Farokh Marvasti
Trading Strategy Stock Movement Prediction Algorithmic Trading Technical Indicator Action Space Reinforcement Learning

September 7, 2022

A Deep Reinforcement Learning Strategy for UAV Autonomous Landing on a Platform
Z. Jiang, G. Song
Reinforcement Learning Diverse Platform Autonomous Landing Action Space Reinforcement Learning