Resource Allocation in Airborne Surveillance Radar

Radars equipped with an Electronically Scanned Antenna (ESA) have the potential of directing the electromagnetic radar beam without mechanically adjusting the antenna. Furthermore, the beam can be redirected instantaneously towards any location in space. Hence, the mechanical constraint of a traditional antenna, which imposes the typically revolving measurement pattern, is relaxed. This means, for example, that targets can be observed in any order. The radar resource management problem is now to design a measurement policy, which optimally utilizes the ESA potentials. This problem is a highly complicated, stochastic optimization problem, and in existing work, many of the practical aspects from real-life are disregarded. For radar designers, practical aspects are important, and more work is needed that focuses on these aspects. In this thesis, optimization-based measurement policies are studied both for offline analysis, and for online beam scheduling in real-time. The general approach in the thesis is to solve a simplified problem denoted as resource allocation, which is posed on a higher abstraction level and on a slower time scale than the original resource management problem. It turns out that the resource allocation problem naturally separates into a series of independent tracking and searching subtasks in the radar system. Among these subtasks, the antenna is regarded as a limited and time shared resource. Given shares of radar resources, i.e., time or energy, each subtask can be optimized locally. Coordination of the subtasks is handled via constraints on the resources, where the constraints are included in a resource allocation algorithm using Lagrange relaxation. Models and performance measures used in optimization are intended to reflect most practical aspects which are important at the considered abstraction level. The resulting algorithms are tested in simulations, showing that optimization-based methods are applicable and useful in design of ESA radar systems.