NN-MPC based Multi-Agent Systems based on FITNET Neural Network Architecture
Main Article Content
Abstract
This paper presents a comparative study of FITNET and CASCADE Neural Network-based Model Predictive Control (NN-MPC) for Multi-Agent Systems (MAS). The NN-MPC framework is applied to the leader-follower consensus problem, where agents coordinate to achieve collective behavior. In this approach, the MPC is utilized to predict the future values the control objective which is optimized by minimizing a cost function with various neural networks architectures. A comparative analysis is conducted based on varying training strategies, and the effectiveness of both the configurations are evaluated through simulation studies on a quadrotor fleet. The results demonstrate that FITNET NN based MPC significantly enhances the performance and consensus behavior of multi-agent systems.