A Parametric Study of CPN’s Convergence Process

The Cognitive Packet Network routing algorithm is a routing algorithm for “self-aware” networks, which continuously monitors the state of the network and is able to respond to changes in network conditions with low latency. In particular, the monitoring and exploration process can be guided by Random Neural Networks to provide the best performance for the lowest search overhead. CPN and RNN have been the focus of several research papers, however these provide little to no detail on how parameters are set. This paper attempts to bridge this gap in the literature by proposing a bench-test experiment of CPN’s initial knowledge gathering process (convergence), whilst modifying the values assigned to key parameters. We discover that one of the parameters controls CPN’s tendency to either produce low-quality results very quickly, but with little improvement over time; or a “slow-but-steadily improving” solution. We also find that another parameter can save some search overhead with minimal impact on the resulting paths’ quality.