Budget-Constrained Digital Twin Synchronization and Its Application on Fidelity-Aware Queries in Edge Computing

Yuchen Li,Weifa Liang,Zichuan Xu,Wenzheng Xu,Xiaohua Jia

DOI: https://doi.org/10.1109/tmc.2024.3455357

IF: 6.075

2024-01-01

IEEE Transactions on Mobile Computing

Abstract:With the advance of mobile edge computing (MEC) and the Internet of Things (IoT), digital twin (DT) has become an emerging technology for provisioning IoT services between the real world and the cyber world. In this paper, we consider the state updating of DTs in an MEC network through synchronizing DTs with their physical objects, by proposing a DT state synchronization framework and synchronization algorithms. We make use of an energy-constrained UAV for data collection in a sensor network, as an illustrative example for the DT state updating of each object (sensor), and then use the DT data of objects (sensors) later for fidelity-aware query services. To this end, we first formulate a novel DT state staleness minimization problem with the aim to minimize the average DT state staleness of all objects for a finite time horizon, under a given update budget per update round. We then propose an optimal algorithm for a special case of the problem where the budget per update round is exactly $K$ objects synchronizing with their DTs. Inspired by the solution of this special case, we then devise an algorithm for the DT state staleness minimization problem by reducing to the award collection maximization problem, assuming that the volume of the update data generated by each object per update round is given. Otherwise, we adopt a deep learning method to predict the volume of the update data. To demonstrate the importance of the DT state staleness in practical applications, we consider fidelity-aware query services in the MEC network where the fidelity of the query result is determined by its source data freshness (or the DT state staleness of querying objects), and we develop a cost-effective evaluation plan for each query. We finally evaluate the performance of the proposed algorithms through simulations. Simulation results demonstrate that the proposed algorithms are promising, where the average DT state staleness by the proposed algorithms are 44.92% less than the baselines, and the query evaluation cost is 19.42% less than the cost delivered by the baseline algorithm.