Abstract: In this paper, we are concerned with the problem ofassigning track tasks, with uncertain processing costsand negligible communication costs, across a set ofhomogeneous processors within a distributed computingsystem to minimize workload imbalances. Since the taskprocessing cost is uncertain at the time of taskassignment, we propose several fast heuristic solutionsthat are extensible, incur very little overhead, andtypically react well to changes in the state of theworkload. The primary differences between the taskassignment algorithms proposed are: (i) the definitionof a task assignment cost as a function of past,present, and predicted workload distribution, (ii)whether or not information sharing concerning the stateof the workload occurs among processors, and (iii) ifworkload state information is shared, the reactivenessof the algorithm to such information (i.e., high-pass,moderate, low-pass information filtering). We show, inthe context of a multisensor-multitarget trackingproblem, that using the heuristic task assignmentalgorithms proposed can yield excellent results andoffer great promise in practice. !15
展开▼
