notesum.ai

Published at November 2

Guiding Multi-agent Multi-task Reinforcement Learning by a Hierarchical Framework with Logical Reward Shaping

cs.AI

cs.LO

Released Date: November 2, 2024

Authors: Chanjuan Liu¹, Jinmiao Cong¹, Bingcai Chen¹, Yaochu Jin¹, Enqiang Zhu¹

Aff.: ¹Not specified

Arxiv: http://arxiv.org/abs/2411.01184v1

Refer to caption

Algorithm	random map		adversarial map
Algorithm	maximum reward	average reward	maximum reward	average reward
MHLRS	8.02	3.81	8.0	3.11
FALCON	1.71	-4.66	-3.68	-7.59
I-LPOPL	-6.72	-7.57	-6.48	-6.48
I-DQN-L	-7.6	-8.07	-7.83	-8.34