Distributed autonomy and trade-offs in online multiobject k-coverage

Abstract

In this article, we explore the online multiobject k-coverage problem in visual sensor networks. This problem combines k-coverage and the cooperative multirobot observation of multiple moving targets problem, and thereby captures key features of rapidly deployed camera networks, including redundancy and team-based tracking of evasive or unpredictable targets. The benefits of using mobile cameras are demonstrated and we explore the balance of autonomy between cameras generating new subgoals, and those responders able to fulfill them. We show that higher performance against global goals is achieved when decisions are delegated to potential responders who treat subgoals as optional, rather than as obligations that override existing goals without question. This is because responders have up-to-date knowledge of their own state and progress toward goals where they are situated, which is typically old or incomplete at locations remote from them. Examining the extent to which approaches overprovision or underprovision coverage, we find that being well suited for achieving 1-coverage does not imply good performance at k-coverage. Depending on the structure of the environment, the problems of 1-coverage and k-coverage are not necessarily aligned and that there is often a trade-off to be made between standard coverage maximization and achieving k-coverage.