31 January 2020 in IEEE Transactions on Vehicular Technology
The advent of 5G technologies enables highly anticipated prospect of the internet of things in smart cities. Smart terminals become a necessity with increasing number of applications and tasks. The limited computational resources and constrained battery energy at terminals compel tasks to be offloaded for execution sometimes. For time-sensitive and safety-critical tasks, fog computing is the preferred option instead of cloud computing, since fog computing brings the computing resources to the edge of networks and thus has relatively shorter response latency. In this context, Vehicular fog computing (VFC) that supplements fog computing by contributing vehicular computing resources has been attracting increasing attention in recent years. However, a few of challenging issues regarding task offloading in VFC still need to be addressed. For example, decision making for task allocation and response latency optimization are not straightforward, since on one hand the connection between vehicles and end users is usually short or intermittent; on the other hand the multi-hop inter-vehicle task forwarding is time-consuming and thus prone to packet loss. To address the latency related issue, we in this paper propose a three-layer architecture of VFC to schedule the tasks offloaded from MDs. RSU is endowed with computing facilities and storing resources such that it can act as not only a gateway to the remote cloud center but also a centralized infrastructure responsible for task scheduling among vehicles. A greedy heuristic based scheduling strategy and algorithms are proposed in the next. Last, an extensive series of experiments are carried out to investigate the proposed scheduling strategy. The results show that the proposed scheduling strategy outperforms other well known heuristics based scheduling strategies such as Min-Min and Max-Min and thus has wide prospects.
14 April 2022 in IEEE Transactions on Green Communications and Networking
Non-intrusive load monitoring (NILM) helps disaggregate a household’s main electricity consumption to energy usages of individual appliances, greatly cutting down the cost of fine-grained load monitoring towards the green home vision. To address the privacy concern in NILM applications, federated learning (FL) could be leveraged for NILM model training and sharing. When applying the FL paradigm in real-world NILM applications, however, we are faced with the challenges of edge resource restriction, edge model personalization, and edge training data scarcity. We present FedNILM, a practical FL paradigm for NILM applications at the edge client. Specifically, FedNILM delivers privacy-preserving and personalized NILM services to large-scale edge clients, by leveraging i) collaborative data aggregation through federated learning, ii) efficient cloud model compression via filter pruning and multi-task learning, and iii) personalized edge model building with unsupervised transfer learning. Our experiments on real-world energy data show that FedNILM can achieve personalized energy disaggregation with the state-of-the-art accuracy, while preserving the user privacy.