A Stochastic Resource Allocation and Task Assignment Framework for Mobile Edge Computing

Team Members

  • Prof. Marwan Krunz (PI)

  • Dr. Mingjie Feng (Postdoctoral Researcher)

  • Wenhan Zhang (Ph.D. Student)

Collaborators:

  • Professor Hang Liu, Catholic University of America

  • Dr. Haris Volos, DENSO International America

  • Professor Yong Xiao, Huazhong University of Science and Technology

Summary

Mobile edge computing (MEC) is an emerging platform for supporting rapidly growing low-latency mobile applications and Internet of Things (IoT) systems. It involves deploying nodes with computing, storage, and communication capabilities at the edge of mobile networks and utilizing client mobile devices for computing and data processing services near end users. This project addresses the challenges of resource allocation and task assignment in MEC systems, enabling horizontal and vertical cooperation among edge nodes, remote cloud data centers, and mobile users for performance optimization of multiple services with different quality of service (QoS) requirements. Specifically, a stochastic framework for cooperative resource allocation and task assignment in MEC systems is designed using deep learning, stochastic game, and optimization techniques.

                

                                                              (a)                                                                                                                      (b)
Figure 1: (a) System architecture of an edge/cloud infrastructure for V2X applications with multiple radio interfaces.​ (b) Dynamic association and edge/cloud offloading in V2X under multiple radio access technologies.

The agenda includes several tasks. First, the complex interactions among the various entities of an MEC system will be investigated, new optimization problems will be formulated, and novel centralized and distributed algorithms will be developed. These algorithms enable a service provider to optimally orchestrate its computational tasks under statistical and time-varying information of user task arrivals and network states. Lightweight heuristic algorithms will also be designed and compared with the optimal solutions. Second, network slicing and resource allocation schemes will be developed so as to allow multiple service providers to dynamically share a virtualized heterogeneous MEC network infrastructure and jointly optimize resource utilization and task assignment for different services. Third, an experimental testbed will be instrumented and used to validate through proof-of-concept prototyping the algorithms developed under this project.

Related Publications

  1. Mingjie Feng, Marwan Krunz, and Wenhan Zhang, “Latency-optimal task partitioning and user association for computation offloading in mobile edge computing networks,” submitted to the IEEE Transactions on Vehicular Technology (TVT), July 21, 2020.
  2. Rong Xia, Yong Xiao, Yingyu Li, Marwan Krunz, and Dusit Niyato, "A generative learning approach for spatio-temporal modeling in connected vehicular networks," Proc. of the IEEE ICC Conference - Wireless Communications Symposium, Dublin, Ireland, 2020.
  3. Wenhan Zhang, Mingjie Feng, Marwan Krunz, and Haris Volos, "Latency prediction for delay-sensitive V2X applications in mobile cloud/edge computing systems," accepted to appear in the Proc. of the IEEE GLOBECOM 2020 Conference, Dec. 2020.
  4. Yong Xiao and Marwan Krunz, “AdaptiveFog: A Modelling and Optimization Framework for Fog Computing in Intelligent Transportation Systems,” submitted to the IEEE Transactions on Mobile Computing (TMC), June 11, 2020.
  5. Yong Xiao, Marwan Krunz, Haris Volos, and Takashi Bando, "Driving in the fog: Latency measurement, modeling, and optimization of LTE-based fog computing for smart vehicles,"  Proc. of the IEEE SECON 2019 Conference, Boston, June 2019.