About Me
❗❗❗LOOKING FOR JOB OPPORTUNITIES❗❗❗
Yuncong Hong (洪云聪) received his PhD in Computer Science from The University of Hong Kong in July 2024, supervised by Prof. Francis C.M. Lau and Prof. Rui Wang. He received his B.Eng from SUSTech in Communication Engineering in 2018.
His research interests include optimization problems in edge computing systems and vehicular networks.
His programming skills are listed but not limited in the following areas: Python, Rust, C/CPP, Nodejs, TypeScript, and etc.
Selected Publications
[1] Y. Hong, B. Lv, R. Wang, H. Tan, Z. Han, H. Zhou, F. C. M. Lau, ``Online Distributed Job Dispatching with Outdated and Partially-Observable Information,’’ in Proc. IEEE MSN, 2020 Best Track Paper Award.
[2] Y. Hong, B. Lv, R. Wang, H. Tan, Z. Han, F. C. M. Lau, ``Distributed Job Dispatching in Edge Computing Networks with Random Transmission Latency: A Low-Complexity POMDP Approach,’’ in IEEE Internet of Things Journal, doi: 10.1109/JIOT.2021.3103798.
[3] Y. Hong, B. Lv, S. Wang, H. Tan, F. C. M. Lau, ``FedCars: A Data-Driven Uplink Scheduling Framework for Efficient In-Vehicle Federated Edge Learning,’’ To be appear in IEEE Transactions on Vehicular Technology, 2024.
[4] Z. Zhang, S. Wang, Y. Hong, L. Zhou and Q. Hao, “Distributed Dynamic Map Fusion via Federated Learning for Intelligent Networked Vehicles,” 2021 IEEE International Conference on Robotics and Automation (ICRA), Xi’an, China, 2021, pp. 953-959, doi: 10.1109/ICRA48506.2021.9561612.
[5] S. Wang, Y. Hong, R. Wang, Q. Hao, Y. -C. Wu and D. W. K. Ng, “Edge Federated Learning via Unit-Modulus Over-The-Air Computation,” in IEEE Transactions on Communications, vol. 70, no. 5, pp. 3141-3156, May 2022, doi: 10.1109/TCOMM.2022.3153488.
Research Experience
Edge Computing
theory
November 2018 - June 2024
Edge Computing in Distributed and Cooperation System
3 published, 0 patent.
In the published papers, we use POMDP to formulate a cooperative distributed job dispatching problem in an edge computing system. We propose an efficient and performance-guaranteed distributed solution framework, which is scalable to the size of the edge computing system.
Vehicular Network and Federated Learning
theory system
November 2019 - June 2024
Online Semi-supervised Federated Learning on High-fidelity Simulators
3 published, 1 ongoing, 1 patent.
In the published papers, we formulate a semi-supervised federated learning problem where multiple vehicles are equipped with a SECOND network for 3D object detection. Firstly, the vehicles would locally infer the objects with a pre-trained model. Then the information of all the inferred objects is uploaded for global fusion. Finally, the vehicles get the distributed global objects list, and try to improve the local network with the previous samples. We implemented the experiment based on a high-fidelity autonomous driving simulator CARLA, and generated some datasets conformed to the KITTI dataset format. The source code is available on Github. One patent is in the discloure stage: 仿真数据生成方法、设备及介质 (CN115577545A).
In the ongoing paper, we formulate a resource allocation problem in the above scenario, to achieve both energy and latency efficiency in online federated training. We note that the randomness of vehicles’ locations would affect both allocation of communication and computation power. Moreover, the sharing of wireless resources further increases the complexity of solution.
VLC-WiFi Integrated System
system
September 2017 - December 2018
VLC Platform over NI Instruments and Out-of-shelf Wi-Fi NIC
0 published, 0 ongoing, 1 patent.
In this project, we implemented a hybrid VLC-WiFi communication system, where the VLC link bears the high-throughput downlink, and Wi-Fi link servers as uplink. We implemented the system on NI instruments with out-of-shelf Wi-Fi NIC, and the retransmission mechanism is implemented on IP layer. The source code of the whole workable system implementation is provided here. The granted patent is 数据传输方法、装置、设备及存储介质 (CN110429979B).
Projects
LaTeX Overleaf VS Code Extension
Open Overleaf (ShareLatex) projects in VSCode, with full collaboration support.
The full user guide is available at GitHub Wiki.
Automation Utility
Single Python file, to distribute tasks and collect execution results or exception traceback on remote clients in cluster. The timing of automation pipeline is based on local system time, which guarantees minimum extra communication cost.
Together with the stream-replay utility and wlsops-hack kernel module, the demo below shows how we automate a Wi-Fi network traffic optimization procedure.
mmap procfs mac80211
This project aims at providing the fastest access to the wireless NIC driver from userspace (the most recommended NIC for now is the intel 8260/9260 series).
In fact, Linux mac80211 subsystem provides a set of APIs to monitor and modify the parameters related to IEEE 802.11 MAC design.
However, the existing tools rely on either ioctl or genl which could not provide real-time and robust access to kernel driver.
In this project, we design one kernel module to hijack the function entry of the driver, and use mmap to establish a shared-memory communication between kernel and userspace.
The experiment results show that we could alter a set of parameters (related to channel access priority) within 10 millisecond for 1000 accesses.
Linux Netfilter Linux Netlink AES-128
This project implements an in-stack network data encryption trial based on Linux Netfilter subsystem.
It uses asynchronous encryption method AES-128 provided by kernel function, to encrypt the payload of L3 (i.e., content of an IP packet) and decrypt it correspondingly at the receiver’s side.
In the possible future work, this project would be extended with an identification mechanism which allows the tx/rx to negotiate the enable of en(de)cryption, and a userspace tool based on Linux Netlink subsystem for key management.