Improved Reinforcement Computing to Implement AntNet-Based Routing Using General NPs for Ubiquitous Environments

摘要

In the ubiquitous convergence era, the traffic managements and quality of services will be made much of a role. Because traditional routing mechanisms are lacking scalability and adaptability, a kind of adaptive routing algorithm called AntNet has attracted the attention. AntNet is an adaptive agent-based routing algorithm that imitates the activities of the social insect. In AntNet, there are implementation constraints due to complex arithmetic calculations for determining a reinforcement value. Besides, a housekeeping core in network processors will be overwhelmed by increasing routing workload for a processing of agents. In this paper, we propose a new reinforcement computing algorithm to overcome these problems. This can be implemented efficiently on packet forwarding engines of conventional network processors. The simulation results show that the proposed AntNet is more adaptive and effective in the performance of the implementation than the original AntNet.