Step Reinforcement Learning

Step reinforcement learning (RL) refines traditional RL by focusing on the granular details of each action step within an episode, rather than solely on the overall episode outcome. Current research emphasizes improving efficiency and addressing challenges like sparse rewards and computationally expensive control loops through techniques such as multi-step updates, variable time-step control, and adaptive reward schemes, often implemented using actor-critic architectures. This approach leads to more efficient exploration, smoother control trajectories, and better performance in various applications, including robotics, dialogue systems, and medical treatments like blood glucose control, ultimately advancing the capabilities and applicability of RL.

Papers

June 29, 2024

Variable Time Step Reinforcement Learning for Robotic Applications
Dong Wang, Giovanni Beltrame
Traditional Reinforcement Learning Robotic Application Discrete Control Step Reinforcement Learning

June 20, 2024

Rewarding What Matters: Step-by-Step Reinforcement Learning for Task-Oriented Dialogue
Huifang Du, Shuqin Li, Minghao Wu, Xuejing Feng, Yuan-Fang Li, Haofen Wang
Reinforcement Learning Task Oriented Response Generation User Satisfaction Dialogue Policy Step Reinforcement Learning

June 3, 2024

MOSEAC: Streamlined Variable Time Step Reinforcement Learning
Dong Wang, Giovanni Beltrame
Traditional Reinforcement Learning Action Duration Adaptive Incentive Step Reinforcement Learning

March 12, 2024

An Improved Strategy for Blood Glucose Control Using Multi-Step Deep Reinforcement Learning
Weiwei Gu, Senquan Wang
Deep Reinforcement Learning Step Prediction Simple Strategy Blood Glucose Glucose Control Step Reinforcement Learning

January 21, 2024

Open the Black Box: Step-based Policy Updates for Temporally-Correlated Episodic Reinforcement Learning
Ge Li, Hongyi Zhou, Dominik Roth, Serge Thilges, Fabian Otto, Rudolf Lioutikov, Gerhard Neumann
Reinforcement Learning Black Box Episodic Reinforcement Learning Binary Forward Exploration Policy Update Step Reinforcement Learning

October 18, 2022

Deep Black-Box Reinforcement Learning with Movement Primitives
Fabian Otto, Onur Celik, Hongyi Zhou, Hanna Ziesche, Ngo Anh Vien, Gerhard Neumann
Reinforcement Learning Deep Reinforcement Learning Motion Primitive Non Markovian Reward Step Reinforcement Learning