.png){kind=icon}
.png){kind=icon}
Return
Efficient Reward Shaping for Multiagent Systems
.png){kind=icon}
PRE
AI
DOI：10.1109/TCNS.2024.3401000.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In this article, we address the reward-shaping problem of large-scale multiagent systems (MASs) using inverse reinforcement learning (IRL). The learning MAS does not have prior knowledge of the cost function of the target MAS and aims to reconstruct it based on the target's demonstrations. We propose a scalable model-free IRL algorithm for a large-scale MAS, where dynamic mode decomposition (DMD) extracts dynamic modes and builds a projection matrix. This significantly reduces the data required while retaining the system's essential dynamic information. The proofs of the algorithm's convergence, stability, and nonuniqueness of the state reward weight are presented. The efficacy of our method is validated with a large-scale consensus network, by comparing the required data sizes and computational time for reward shaping with and without DMD.
Keywords:
Artificial neural networks
Heuristic algorithms
Optimal control
Control systems
Network systems
Dimensionality reduction
Stability criteria
Data-driven control
dynamic mode decomposition (DMD)
inverse reinforcement learning (IRL)
large-scale system
optimal control
Journal
IF:
5
Papers:
1.6K
Citations:
5.8K