Research on Load Balancing Based on Multi-agent in Ubiquitous Networks

Next-generation Network wireless Communication System has a variety of access modes. In order to solute how to select a proper access network under this feature and to realize the load balancing, A ubiquitous networks configuration based on multi-agent is proposed in this paper, use multi-agent Q-learning to realize the load balancing in Ubiquitous Networks with heterogeneous Radio Access Technologies (RAT). Agent in the Radio Access Network (RAN) collects radio access networks' loading information, then feed back them by communicate between agents. Agents get optimal policy through trial-and-error and interaction with wireless environment and learn to allocate the proper RAN for each session. Simulation results show that the proposed algorithm realizes the autonomy of resource control through the agents' online learning process and achieves the load balancing through the agents' coordinate.