Guifu Ma

Dr. Guifu Ma

orcid

Research Focus & Metrics

Academic Career Length

5Years (2022-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
28 November 2022 in IEEE Transactions on Intelligent Vehicles
Verified
Journal Article
Closed access
GMGuifu Ma
Guifu Ma
XWXiaowei Wang
Xiaowei Wang
MHManjiang Hu
Manjiang Hu
WOWenjie Ouyang
Wenjie Ouyang
XCXiaolong Chen
Xiaolong Chen
YLYang Li
Yang Li
Computation offloading is considered as a promising method to improve computation performance of Intelligent Vehicles (IVs), where IVs can offload resource-hungry applications to Mobile Edge Computing (MEC) servers. However, due to the dynamic environment of Intelligent Transportation Systems, it's challenging to develop an effective offloading scheme that ensures stable operation of the system. Furthermore, IVs only offload tasks to MEC servers or parked vehicles, how to take full advantage of resources in a random traffic flow is a main challenge. In this paper, we propose a vehicular-cloud-assisted MEC network in urban cities to fully explore the idle resources and ensure the stability of task queue. Specifically, we develop a novel virtual platform based on deep neural network that predicts vehicles trajectory and creates vehicular cloud to integrate computation resources. We then design a vehicular-cloud-assisted offloading scheme that aims to maximize task throughput in continuous time slots subject to multi-constraint. To address the coupling among multiple variables and different time slots, we put forward a lightweight framework to apply the Lyapunov optimization in combination with deep reinforcement learning. In particular, we apply Lyapunov drift-plus-penalty approach to decompose the initial problem into per-slot subproblems and control the long-term task queue stability. To further reduce the computation complexity, we decouple the subproblem into three processes, and the deep Q-network is put forward to solve it. Extensive simulations demonstrate that under different task arrivals, our proposed scheme can achieve the task queue stability within 10 time slots with a low transmission power.

6 May 2024 in IEEE Transactions on Intelligent Transportation Systems
Verified
Journal Article
Closed access
GMGuifu Ma
Guifu Ma
MHManjiang Hu
Manjiang Hu
XWXiaowei Wang
Xiaowei Wang
HLHaoran Li
Haoran Li
YBYougang Bian
Yougang Bian
DWDi Wu
Di Wu
In the foreseeable Intelligent Transportation System, Intelligent Connected Vehicles (ICVs) will play an important role in improving travel efficiency and safety. However, it is challenging for ICVs to support the resource-hungry autonomous driving applications due to the limitation of hardware computing power. Fortunately, the emergence of Multi-access Edge Computing helps overcome this limitation effectively. This paper addresses the vehicle-to-edge server computation offloading conundrum by optimizing the trade-offs in partial offloading and resource allocation. Proposing a distributed approach, this study confronts the multi-variable non-convex challenge directly by decoupling variables and deriving constraint-based bounds that guide the decisions for offloading and allocation. A novel low-complexity distributed algorithm is introduced that not only tends toward optimal but also demonstrates superior real-time applicability and efficiency, illustrated through enhanced performances both in simulated trials and genuine vehicular edge computing settings. The algorithm’s practical effectiveness addresses a notable gap between the theoretical models for computation offloading and actual real-life execution, reinforcing the soundness and relevance of the proposed method. Furthermore, its advanced integration with federated learning frameworks marks a leading-edge application, substantiating significant enhancements in computational efficiency and robustness.

9 July 2025 in Sensors
Verified
Journal Article
Open access
PDF file
GMGuifu Ma
Guifu Ma
SJSeyed Mohamad Javidan
Seyed Mohamad Javidan
Yiannis AmpatzidisYiannis Ampatzidis
Yiannis Ampatzidis
ZZZhao Zhang
Zhao Zhang
Accurate and early detection of plant diseases is essential for effective management and the advancement of sustainable smart agriculture. However, building large annotated datasets for disease classification is often costly and time-consuming, requiring expert input. To address this challenge, this study explores the integration of few-shot learning with hyperspectral imaging to detect four major fungal diseases in tomato plants: Alternaria alternata, Alternaria solani, Botrytis cinerea, and Fusarium oxysporum. Following inoculation, hyperspectral images were captured every other day from Day 1 to Day 7 post inoculation. The proposed hybrid method includes three main steps: (1) preprocessing of hyperspectral image cubes, (2) deep feature extraction using the EfficientNet model, and (3) classification using Manhattan distance within a few-shot learning framework. This combination leverages the strengths of both spectral imaging and deep learning for robust detection with minimal data. The few-shot learning approach achieved high detection accuracies of 85.73%, 80.05%, 90.33%, and 82.09% for A. alternata, A. solani, B. cinerea, and F. oxysporum, respectively, based on data collected on Day 7 post inoculation using only three training images per class. Accuracy improved over time, reflecting the progressive nature of symptom development and the model’s adaptability with limited data. Notably, A. alternata and B. cinerea were reliably detected by Day 3, while A. solani and F. oxysporum reached dependable detection levels by Day 5. Routine visual assessments showed that A. alternata and B. cinerea developed visible symptoms by Day 5, whereas A. solani and F. oxysporum remained asymptomatic until Day 7. The model’s ability to detect infections up to two days before visual symptoms emerged highlights its value for pre-symptomatic diagnosis. These findings support the use of few-shot learning and hyperspectral imaging for early, accurate disease detection, offering a practical solution for precision agriculture and timely intervention.

2 October 2023 in IEEE Systems Journal
Verified
Journal Article
Closed access
XWXiaowei Wang
Xiaowei Wang
HLHaoran Li
Haoran Li
MHManjiang Hu
Manjiang Hu
QDQuanli Dou
Quanli Dou
WOWenjie Ouyang
Wenjie Ouyang
HQHongmao Qin
Hongmao Qin
High-definition (HD) maps can provide an accurate representation of the road ahead and information on the surrounding environment, which is the key to the success of autonomous driving. Open-pit mine is a typical unstructured road environment with dynamic and varying terrain, posing huge challenges to the environmental mapping. This study proposes an HD map construction and update system for autonomous driving in open-pit mines. First, we analyze the map elements of autonomous driving and the environmental characteristics of open-pit mines. According to the types, uses, and update frequency of map elements, we establish a multimodal and multilevel map structure to realize the all-round unified map description. Second, in order to construct and update various map elements in the map structure, we propose a map construction method by the fusion of localization and perception and a multiscene map update method by analyzing the updated scenes of open-pit mines. Then, based on the proposed map structure and method, we integrate the software and hardware of the map platform to design an HD map construction and update system for open-pit mines. Finally, tests are performed on the unstructured road scenes to verify the performance of the proposed scheme by comparing the traditional road network map and the point cloud map. Field tests demonstrate that in the same site, compared with the point cloud map, the proposed method reduces the map data storage space by 98% and achieves the accuracy and real-time performance of map update with a lower delay within the error range of 0.15 m.

22 April 2024 in IEEE Transactions on Intelligent Vehicles
Verified
Journal Article
Closed access
GMGuifu Ma
Guifu Ma
YBYougang Bian
Yougang Bian
HQHongmao Qin
Hongmao Qin
CYChenlong Yin
Chenlong Yin
CCChaoyi Chen
Chaoyi Chen
KLKeqiang Li
Keqiang Li
Enhancing autonomous driving through Federated Learning (FL) in Intelligent Connected Vehicles (ICVs) confronts challenges like limited scalability of central management, computational strains on diverse ICVs, and inefficiencies due to stragglers. This paper presents Advance-FL, a deep reinforcement learning based Adaptive asynchronous online Federated Learning with a computation offloading assisted framework for vehicular edge cloud computing networks to mitigate the above challenges. Innovatively, Advance-FL incorporates the concept of “straggler rate”, a metric originally introduced in this study to quantify the degree of lag in participant computation and training, thus enabling targeted mitigation strategies. By employing an asynchronous advantage actor-critic approach for adaptive data offloading and dynamic local iteration adjustments, Advance-FL effectively alleviates computation resource shortages and harmonizes the balance between model accuracy and straggler impact by dynamically managing the straggler rate. Critical findings include over 62% reduction in training times and a 2% ∼ 9% increase in model accuracy across varying non-IID data scenarios and reducing training time by more than 6 times as the number of ICVs increases, compared to prevailing methods. Additionally, experiments in both static and dynamic test-bed further validate Advance-FL's superior scalability and robustness over state-of-the-art approaches, particularly in maintaining high performance under straggler effects, and showcasing robust adaptability across long-term operations, large-scale datasets and abnormal situations

27 November 2024 in IEEE Internet of Things Journal
Verified
Journal Article
Closed access
XWXiaowei Wang
Xiaowei Wang
WOWenjie Ouyang
Wenjie Ouyang
YLYang Li
Yang Li
CYChenlong Yin
Chenlong Yin
MCMingyang Chen
Mingyang Chen
YHYingbai Hu
Yingbai Hu
Proactive caching scheme proactively caches data before the request comes, which is essential for optimizing of the resource utilization and ensuring the stability of cached contents. However, current proactive caching methods suffer from high computational complexity and often fall into suboptimal solutions. This study proposes a hierarchical reinforcement learning (HRL)-based proactive caching scheme to determine whether to update the caching queue and how to update the caching queue. Specifically, we build a bi-level optimization model to maximize the content cache hit ratio while minimizing the content cache cost. We formulate the upper-level and lower-level optimization problems as Markov decision process (MDP) models and employ double deep Q-learning methods to solve the optimal caching strategy. In particular, the content history request features and the user preference generated by the collaborative filtering-based recommendation system are used as the state inputs of caching decisions. In addition, we design a two-stage curriculum learning paradigm for training the agents from easy tasks to hard tasks. At the first stage, a greedy decision tree with a posterior knowledge is designed to provide the upper-level decisions for training the lower-level agent in the easy task. Then, we train all agents together in the hard task while freezing the trained lower-level agent. We conduct both the simulation and on-road experiments to validate the effectiveness of our method, and results indicate that our method outperforms the baseline methods by 2.2%–37.2% in terms of cache hit ratio and 75.4%–77.6% in terms of cache cost.