Battery-Aware Scheduling of Mixed Criticality Systems.

Wireless systems such as satellites and sensor networks are often battery-powered. To operate optimally they must therefore take the performance properties of real batteries into account. Additionally, these systems, and therefore their batteries, are often exposed to loads with uncertain timings. Mixed criticality and soft real-time systems may accept deadline violations and therefore enable trade-offs and evaluation of performance by criteria such as the number of tasks that can be completed with a given battery. We model a task set in combination with the kinetic battery model as a stochastic hybrid system and study its performance under battery-aware scheduling strategies. We believe that this evaluation does not scale with current verification techniques for stochastic hybrid systems. Instead statistical model checking provides a viable alternative with statistical guarantees. Based on our model we also calculate an upper bound on the attainable number of task instances from a battery, and we provide a battery-aware scheduler that wastes no energy on instances that are not guaranteed to make their deadlines.