Curiosity Driven Reinforcement Learning

Curiosity-driven reinforcement learning (CDRL) enhances standard reinforcement learning by incorporating an intrinsic reward mechanism that motivates agents to explore novel states and actions, accelerating learning and improving performance. Current research focuses on integrating CDRL with various architectures, including those based on prediction error and semantic similarity, to address challenges in diverse applications such as dialogue systems, robotics control (e.g., humanoid character animation and quadrotor flight), and text-based game playing. This approach shows promise in improving sample efficiency, robustness, and the ability to learn complex behaviors in challenging environments, ultimately advancing both theoretical understanding and practical applications of reinforcement learning.

Papers

January 31, 2024

Enhancing End-to-End Multi-Task Dialogue Systems: A Study on Intrinsic Motivation Reinforcement Learning Algorithms for Improved Training and Adaptability
Navin Kamuni, Hardik Shah, Sathishkumar Chintala, Naveen Kunchakuri, Sujatha Alla Old Dominion
Study Feature Adaptive Importance Intrinsic Reward Improved Training Downstream Dialogue Task Curiosity Driven Reinforcement Learning

August 14, 2023

Neural Categorical Priors for Physics-Based Character Control
Qingxu Zhu, He Zhang, Mengting Lan, Lei Han
Reinforcement Learning Physic Based Motion Prior Vector Quantized Variational Specific Prior Curiosity Driven Reinforcement Learning

July 28, 2023

Curiosity-Driven Reinforcement Learning based Low-Level Flight Control
Amir Ramezani Dooraki, Alexandros Iosifidis
Flight Control Curiosity Inducing Situation Independent Learning Directed Exploration Artificial Curiosity Curiosity Driven Reinforcement Learning

March 26, 2022

Aggressive Quadrotor Flight Using Curiosity-Driven Reinforcement Learning
Qiyu Sun, Jinbao Fang, Wei Xing Zheng, Yang Tang
Reinforcement Learning Algorithm Curiosity Driven Reinforcement Learning Quadrotor Flight

March 20, 2022

Perceiving the World: Question-guided Reinforcement Learning for Text-based Games
Yunqiu Xu, Meng Fang, Ling Chen, Yali Du, Joey Tianyi Zhou, Chengqi Zhang
Reinforcement Learning Deep Reinforcement Learning World Event Perception Module Text Based Game Two Stage Training Curiosity Driven Reinforcement Learning

March 5, 2022

Target Network and Truncation Overcome The Deadly Triad in $Q$-Learning
Zaiwei Chen, John Paul Clarke, Siva Theja Maguluri
Linear Function Approximation Function Approximation Q$ Learning Finite Sample Guarantee Goal Node Balanced Truncation Curiosity Driven Reinforcement Learning