Stochastic Steepest-Descent Optimization of Multiple-Objective Mobile Sensor Coverage
In this paper, the authors propose a steepest descent method to compute control parameters for achieving an optimal tradeoff between multiple objectives in stateless stochastic scheduling. By stateless, they mean that the scheduling algorithm does not need any bookkeeping of prior service received by clients, but the scheduling decision is determined entirely by a constant-time coin toss operation. The notion of optimal balance between multiple performance objectives is user- or application-defined. For example, some applications may give a larger importance to rate than fairness, whereas other applications may require the opposite. Their ability to support a wide range of multi-objective performance tradeoffs and optimize the tradeoff is absent in existing stochastic scheduling algorithms.