Hide/Show Apps

On optimal resource allocation in phased array radar systems

Ircı, Ayhan
In this thesis, the problem of optimal resource allocation in real-time systems is studied. A recently proposed resource allocation approach called Q-RAM (Quality of Service based Resource Allocation Model) is investigated in detail. The goal of the Q-RAM based approaches is to minimize the execution speed in real-time systems while meeting resource constraints and maximizing total utility. Phased array radar system is an example of a system in which multiple tasks contend for multiple resources in order to satisfy their requirements. In this system, multiple targets are tracked (each a separate task) by the radar system simultaneously requiring processor and energy resources of the radar system. Phased array radar system is considered as an illustrative application area in order to comparatively evaluate the resource allocation approaches. For the problem of optimal resource allocation with single resource type, the Q-RAM algorithm appears incompletely specified, namely it does not have a termination criteria set that can terminate the algorithm in all possible cases. In the present study, first, the Q-RAM solution approach to the radar resource allocation problem with single resource type is extended to give a global optimal solution in all possible termination cases. For the case of multiple resource types, the Q-RAM approach can only generate near-optimal results. In this thesis, for the formulated radar resource allocation problem with multiple resource types, the Methods of Feasible Directions are considered as an alternative solution approach. For the multiple resource type case, the performances of both the Q-RAM approach and the Methods of Feasible Directions are investigated in terms of optimality and convergence speed with the help of Monte-Carlo simulations. It is observed from the results of the simulation experiments that the Gradient Projection Method produce results outperforming the Q-RAM approach in closeness to optimality with comparable execution times.