Distributed autonomous control of concurrent combat tasks

This article discusses the use of decentralized model predictive control to manage a set of related combat tasks (e.g., the destruction of anti-aerial defenses in a given geographical area by a wing of uninhabited air vehicles (UAVs)). The controller is meant to operate autonomously as it attempts to manage conditions in which success in all its objectives cannot be guaranteed at every step. It is also meant to work in a distributed fashion. This implies that complete knowledge of the state of each task is not needed at the central level. The approach presented suggests a way of aggregating battle information to be used at the higher levels of command. The aggregation used is based on combat models and physical concepts and does not rely on arbitrary valuations of assets and targets.