Directional Bias and Pheromone for Discovery and Coverage on Networks

Natural multi-agent systems often rely on "correlated random walks" (random walks that are biased toward a current heading) to distribute their agents over a space (e.g., for foraging, search, etc.) Our contribution involves creation of a new movement and pheromone model that applies the concept of heading bias in random walks to a multi-agent, digital-ants system designed for cyber-security monitoring. We examine the relative performance effects of both pheromone and heading bias on speed of discovery of a target and search-area coverage in a two-dimensional network layout. We found that heading bias was unexpectedly helpful in reducing search time and that it was more influential than pheromone for improving coverage. We conclude that while pheromone is very important for rapid discovery, heading bias can also greatly improve both performance metrics.