ABSTRACT
The problems in distributed networks, which often appear in complex applications, such as sensor, traffic or logistics networks can be solved using multi-agent system (MAS) architectures. This chapter describes previous work related to MAS applications for solving problems in distributed networks, the Distributed data processing and mining (DDPM) field in MASs, travelling time forecasting models, and formulates the research question. It presents a distributed cooperative recursive forecasting algorithm based on multivariate linear regression. The chapter examines a simple tutorial example to illustrate the proposed technique. It discusses the multivariate kernel-based regression model used for streaming data and proposes a cooperative learning algorithm for optimal forecasting in a distributed MAS architecture. The chapter provides case studies using data from Hanover and a comparison of different MAS architectures and forecasting methods using real-world data. The agents in MAS usually have two important modules: a DDPM module and a distributed decision support module.
