TG-MG: Task grouping based on MDP graph for multi-task reinforcement learning

Rui Pan

doi:10.1016/j.eswa.2026.131922

TG-MG: Task grouping based on MDP graph for multi-task reinforcement learning

2026-03-03

0

PRE

AI

DOI：10.1016/j.eswa.2026.131922

Original

Original request for help

Save

Abstract

En 中文

• TG-MG is the first effort to explore the task grouping problem in the field of MTRL. • We propose an MDP graph learning method to measure MDP similarity. • We propose a novel Task Grouping method based on MDP Graph (TG-MG) for MTRL. • Extensive experiments and comparative analyses verify the effectiveness of TGMG.

Keywords:

Task Grouping

MDP Graph

Multi-Task Reinforcement Learning

Graph Learning

Markov Decision Process

TG-MG: Task grouping based on MDP graph for multi-task reinforcement learning

Abstract

Journal

Organization