Characterizing robustness in dynamic real-time systems

The problem of robust task allocation is motivated by the need to deploy real-time systems in dynamic operational environments. Existing robust allocation approaches employ coarse robustness metrics, which can result in poor allocations. This paper proposes a metric that accurately characterizes a system’s robustness within feasible allocation regions. An allocation algorithm is provided to find allocations that are both feasible and robust; the robustness as measured by the metric is shown to have theoretical bounds. Experiments demonstrate that the algorithm produces good and scalable performance compared with several heuristic algorithms.