1 October 2024 in IEEE Internet of Things Journal
Mobile Edge Computing (MEC) is a promising computing scheme to support computation-intensive AI applications in vehicular networks, by enabling vehicles to offload computation tasks to edge computing servers deployed on Road Side Units (RSUs) that approximate to them. In this work, we consider an MEC-enabled Vehicular Edge Network (VEN), where each vehicle can offload tasks to edge/cloud computing servers via vehicle-to-infrastructure (V2I) links or to other end-vehicles via vehicle-to-vehicle (V2V) links. In such a cloud-edge-end collaborative offloading scenario, we focus on the joint task offloading, scheduling, and resource allocation problem for vehicles, which is challenging due to the online and asynchronous decision-making requirement for each task. To solve the problem, we propose a Multi-layer Deep Reinforcement Learning (DRL) based approach, where each vehicle constructs and trains three modules to make different layers’ decisions: (i) Offloading Module (first layer), determining whether to offload each task, by using the Dueling and Double Deep Q-Network (D3QN) framework; (ii) Scheduling Module (second layer), determining where and how to offload each task in the offloading queues, together with the transmission power, by using the Parameterized Deep Q-Network (PDQN) framework; and (iii) Computing Module (third layer), determining how much computing resource to be allocated for each task in the computation queues, by using classic optimization techniques. We provide the detailed algorithm design and perform extensive simulations to evaluate its performance. Simulation results show that our proposed algorithm outperforms the existing algorithms in the literature, and can reduce the average cost by 25.86%–72.51% and increase the average satisfaction rate by 3.48%-90.53%.
5 February 2024 in IEEE Internet of Things Journal
While computing task offloading in mobile edge computing (MEC) has been extensively studied, existing research has primarily focused on the mobility-awareness arising from opportunistic contact between edge devices. However, in a crowdsourced MEC system, it is essential to incentivize edge users to relocate for offloading tasks to crowdsourced edge devices. This movement-aware task offloading is particularly important for the crowdsourced system operation and has not yet been thoroughly explored. Moreover, the situation becomes more complex when users are socially connected and D2D-enabled, yet few works have considered these characteristics in combination, particularly from an economic incentive perspective. Therefore, a new incentive framework is needed to analyze the economic issues comprehensively. In this work, we focus on designing a truthful incentive mechanism, where socially-connected D2D users can be incentivized to move around for offloading tasks. To truthfully elicit private information, we model the resource allocation between edge devices and users as a multi-seller multi-buyer double auction mechanism with realistic MEC constraints, such as delay and storage limitation. Theoretically, we show that the proposed mechanism is computationally efficient and achieves desirable economic properties, including truthfulness, individual rationality, and budget balance. Simulations demonstrate that the mechanism achieves good system efficiency, with a performance improvement of 20% compared to state-of-the-art baselines.
28 November 2022 in IEEE Transactions on Network Science and Engineering
Mobile edge caching allows edge devices to cache popular contents and deliver them to end-users directly, and hence can effectively alleviate the increasingly heavy backbone loads and improve the quality-of-service of end-users. Peer content sharing enables edge devices to share the cached contents with each other, and can further increase the caching capability and efficiency. While lots of research efforts have been made to edge caching or content sharing, it remains largely open to devise a joint caching and sharing framework and study the complicated technical and economic interplay between both technologies. In this paper, we propose a joint framework for mobile edge caching and peer content sharing, and focus on studying the strategic behaviors and interactions of edge devices in the joint framework. Specifically, we model their interactions as a non-cooperative game, where each edge device ( player ) can choose to be an agent (who caches and shares contents with others) or a requester (who doesn't cache but requests contents from other devices) of each content. We characterize the game equilibrium under a generic usage-based pricing scheme (for content sharing) and analyze its existence and uniqueness systematically. We further design a best response based iterative learning algorithm for players to update their behaviors and reach the equilibrium in a self-enforcing manner. Moreover, we analyze the Price of Anarchy (PoA) of the game equilibrium to measure the system performance degradation due to the selfish strategic behaviors of game players, based on which we further design a pricing scheme (for content sharing) to reduce such performance degradation. Simulations show that our framework can reduce the total system cost by up to 60%, comparing with the benchmark pure caching system without peer content sharing.
17 October 2017 in IEEE Transactions on Mobile Computing
Mobile crowdsensing (MCS) is a promising sensing paradigm that leverages the diverse embedded sensors in massive mobile devices. A key objective in MCS is to efficiently schedule mobile users to perform multiple sensing tasks. Prior work mainly focused on interactions between the task-layer and the user-layer, without considering tasks' similar data requirements and users' heterogeneous sensing capabilities. In this work, we propose a three-layer data-centric MCS model by introducing a new data-layer between tasks and users, enable different tasks to reuse the same data items, hence effectively leverage both task similarities and user heterogeneities. We formulate a joint task selection and user scheduling problem based on the new framework, aiming at maximizing social welfare. We first analyze the centralized optimization problem with the statistical information of tasks and users, and show the bound of the social welfare gain due to data reuse. Then we consider the two-sided information asymmetry of selfish task-owners and users, and propose a decentralized market mechanism for achieving the centralized social optimality. In particular, considering the NP-hardness of the optimization, we propose a truthful two-sided randomized auction mechanism that ensures computational efficiency and a close-to-optimal performance. Simulations verify the effectiveness of our proposed model and mechanism.
28 January 2020 in IEEE Transactions on Cognitive Communications and Networking
With the rapid advance of smart wireless technologies, a plethora of human behavioral data are generated in 5G networks, which is reported capable to improve network performance by leveraging intelligent channel resource allocation through big data analytics. However, what information can be extracted for the network mobility management, how to exploit the knowledge for resource allocation and to meet the user-centric quality of experience (QoE) are not well understood and fully explored. To address this problem, we propose an online learning algorithm for dynamic channel allocation based on contextual multi-armed bandit (CMAB) theory. Especially, we divide the stochastic human behavioral data into two categories: the user location and the QoE-driven context. Noticing that the distributions of CSI vary spatially, we define a set of user’s geographic locations that shares the same set of CSI distributions as a cluster, and the stochastic channel distributions vary across clusters. The problem is formulated as a novel latent SCB problem, where the proposed agnostic SCB algorithm could automatically find the underlying clusters and significantly improve the learning performance. We then extend our online learning algorithm into the practical multi-user random access scenario. We conduct experiments on a real dataset collected from China Mobile, which indicate that our algorithms outperform existing approaches tremendously and perform extremely well in large-scale and high-mobility networks.
19 January 2023 in IEEE/ACM Transactions on Computational Biology and Bioinformatics
The recognition of protein-protein interaction sites (PPIs) is beneficial for the interpretation of protein functions and the development of new drugs. Traditional biological experiments to identify PPI sites are expensive and inefficient, leading to the generation of various computational methods to predict PPIs. However, the accurate prediction of PPI sites remains a big challenge due to the existence of the sample imbalance issue. In this work, we design a novel model that combines convolutional neural networks (CNNs) with Batch Normalization to predict PPI sites, and employ an oversampling technique Borderline-SMOTE to address the sample imbalance issue. In particular, to better characterize the amino acid residues on the protein chains, we employ a sliding window approach for feature extraction of target residues and their contextual residues. We verify the effectiveness of our method by comparing our method with the existing state-of-the-art schemes. The performance validations of our method on three public datasets achieve accuracies of 88.6%, 89.9%, and 86.7%, respectively, all showing improved accuracies compared with the existing schemes. Moreover, the ablation experiment results suggest that Batch Normalization can greatly improve the generalization and the prediction stability of our model.