Understanding Agent Competency: Effects of Environment Complexity on Area Coverage Time

As an increasing number of search-and-rescue (SAR) missions begin to incorporate robotic agents and algorithms to cover search areas, the task of selecting an appropriate algorithm or robot for a specific mission becomes critical. This is particularly challenging since the search time or area coverage performances of SAR algorithms may vary significantly from one operating environment to another. While previous works have evaluated the performance of various SAR algorithms, few have examined how this performance changes as a function of the complexity of the search area or environment. In the presented work, we seek to overcome this knowledge gap by (a) developing the notion of algorithm competency, and (b) generating a competency profile for a specific biologically-inspired SAR algorithm (Lévy walk search) that relates the algorithm's performance to the complexity of the search environment. Our simulation-based study leverages several existing measures of network topological complexity. Our results show that we can successfully generate competency profiles that indicate positive correlation between the area coverage time and certain complexity measures of the environment. We have also found that some algorithms work better in low complexity environments, while other algorithms outperform them in more complex environments, indicating that competency profiles can aid in the selection of appropriate algorithms.