Utility‐Aware Scheduling of Sensor Update Transactions on Multiprocessor Platforms

Scheduling sensor update transactions to maintain data temporal validity has been recognized as an important problem in cyber‐physical system research. Previous work focuses on minimizing the update workload produced by transactions. In this paper, we study the problem of scheduling a set of sensor update transactions on multiprocessor platforms to maximize the overall system utility, while providing deterministically guaranteed temporal validity. We propose a method called utility‐aware partitioned scheduling (UA‐P) to address this problem. UA‐P selects a proper version for each transaction based on a feasibility condition of the transaction set. It then conducts transaction assignment and version adjustment to further increase the system utility. Extensive experiments are conducted. The results show that UA‐P outperforms the baseline methods in terms of system utility.