Tasks allocation for real-time applications in heterogeneous sensor networks for energy minimization

Tasks allocation plays an important role in resource-limited sensor networks. However, existing methods separate energy saving and QoS (Quality-of-Service)-guarantee into two issues, consequently, the scheduling length could be very long which violating user's deadline requirement or might waste lots of energy. This paper discusses the problem of allocating a set of real-time tasks with dependencies onto a heterogeneous sensor network. In order to find an optimal allocation that minimize the overall energy consumption while meeting user's deadline, we exploit the divide-and-conquer technique, first group tasks into tasks partitions and then optimally solve the scheduling problem in branches with several sequential tasks by modeling the branches as a Markov Decision Process. Sensors failure can be handled by rescheduling part of the tasks graph. Experiments results show our proposed method significantly improve the performance of sensor network in terms of energy saving and QoS-guarantee compared with two other heuristic algorithms.