Modern Deep Reinforcement Learning

Modern deep reinforcement learning (DRL) focuses on training agents to make optimal sequential decisions in complex environments using deep neural networks. Current research emphasizes improving sample efficiency through techniques like hierarchical world models and more efficient exploration strategies, as well as addressing challenges in continuous action spaces with methods like Extreme Q-Learning. These advancements are impacting diverse fields, from robotics (e.g., autonomous UAV docking) and animation to optimizing experimental design, by enabling more efficient and effective learning in challenging real-world scenarios. The development of energy-efficient DRL algorithms, inspired by biological systems, is also a growing area of interest.

Papers

June 2, 2024

Deep Reinforcement Learning for Sim-to-Real Policy Transfer of VTOL-UAVs Offshore Docking Operations
Ali M. Ali, Aryaman Gupta, Hashim A. Hashim
Deep Reinforcement Learning Novel Reinforcement Learning Autonomous Landing Vertical Take Off and Landing Modern Deep Reinforcement Learning

October 8, 2023

Hieros: Hierarchical Imagination on Structured State Space Sequence World Models
Paul Mattes, Rainer Schlosser, Ralf Herbrich
Transformer Based World Model Representation Online Matter Modern Deep Reinforcement Learning

January 5, 2023

Extreme Q-Learning: MaxEnt RL without Entropy
Divyansh Garg, Joey Hejna, Matthieu Geist, Stefano Ermon
Offline Reinforcement Learning Q Learning Q Function Stable Entropy Maximum Entropy Reinforcement Learning MaxEnt RL Modern Deep Reinforcement Learning

December 29, 2022

Tuning Synaptic Connections instead of Weights by Genetic Algorithm in Spiking Policy Network
Duzhen Zhang, Tielin Zhang, Shuncheng Jia, Qingyu Wang, Bo Xu
Genetic Algorithm Balancing Weight Spike Protein Synaptic Plasticity Synaptic Connectivity Modern Deep Reinforcement Learning Synaptic Downscaling

March 7, 2022

A Survey on Reinforcement Learning Methods in Character Animation
Ariel Kwiatkowski, Eduardo Alvarado, Vicky Kalogeiton, C. Karen Liu, Julien Pettré, Michiel van de Panne, Marie-Paule Cani
Reinforcement Learning Timely Survey Character Animation Reactive Behavior Physic Based Character Modern Deep Reinforcement Learning

February 2, 2022

Optimizing Sequential Experimental Design with Deep Reinforcement Learning
Tom Blau, Edwin V. Bonilla, Iadine Chades, Amir Dezfouli
Deep Reinforcement Learning Markov Decision Process Design Space Optimal Design Sequential Experimental Design Sequential Experiment Modern Deep Reinforcement Learning