Edge Computing and Edge AI
▶ Introduction

In recent years, wireless users have become producers and consumers of contents as their devices are now embedded with various sensors [1], which help in creating and collecting various types of data from different domains such as energy, agriculture, healthcare, transport, security, and smart homes, among others. Indeed, by the year 2020, it is anticipated that 50 billion things will be connected to the Internet, which is equivalent to 6 devices per person on the planet [2]. Therefore, the devices of wireless users will be anywhere, anytime, and connected to anything [3]. With large-scale interconnection of people and things, there will be a tremendous growth of data traffic with different characteristics (unstructured, quasi-structured, and semi-structured) whose scale, distribution, diversity, and velocity fall into a big data framework that requires big data infrastructure and analytics.

Multi-access Edge Computing: Since the resources (e.g., battery power, CPU cycles, memory, and I/O data rate) of user devices are limited, user devices must offload computational tasks and big data to the cloud [4]. However, for effective big data analytics of delay sensitive and context-aware applications, there is a strong need for low latency and reliable computation. As such, reliance on a cloud can hinder the performance of big data analytics, due to the associated overhead and end-to-end delays [3], [5]. To reduce end-to-end delay and the need for extensive user cloud communication, Multi-access Edge Computing (MEC) has been introduced by the European Telecommunications Standards Institute (ETSI) as a supplement to cloud computing and mobile edge computing [6]. MEC extends cloud computing capabilities by providing IT-based services and cloud computing capabilities at the networks edges. In other words, MEC pushes 4C (Computing, Caching, Communication, or Control) to the edge of the network [7]. Typically, MEC servers are deployed at the Base Stations (BSs) of a wireless network (e.g., a cellular network) for executing delay sensitive and context-aware applications in close proximity to the users [8], [9], [10].

Edge Computing and Artificial Intelligence: Artificial Intelligence (AI) technology, especially centralized AI, is currently being applied in various sectors such as agriculture, financial, retail, and energy sector. The centralized AI, also known as cloud-AI, is implemented at the cloud data center or high-end computing machine, where it can get the full access to the globally collected data, the massive amount of storage and computing power. The cloud-AI based solution consumes a large amount of bandwidth to transfer raw data to the cloud data center as well as sending the raw data to the cloud has privacy, security, and legal issues. Currently, due to the growth of intelligent devices to provide real-time services with low latency and must operate under high reliability, even when network connectivity is lost. In here, edge AI plays the main role to solve the aforementioned problems, where edge AI is the combination of edge computing with various types of machine learning solutions. The pros and cons of cloud AI and Edge AI are describing in Fig. 1. Note that the edge-AI will not replace the cloud-AI and two approaches will complement each other.

Figure 1 Cloud AI vs. Edge AI

▶ Research Challenges and Issues

  • MEC server resources are limited compared to the remote cloud [12]. Therefore, when each MEC server operates independently, it cannot efficiently handle big data stemming from users’ devices and significantly relieve the data exchange between users’ devices and the remote cloud. Therefore, to reduce the delay, cooperation among MEC servers for resource sharing and optimization of the resource utilization are needed.
  • The integration of MEC with a mobile network environment raises a number of challenges related to the coordination of both MEC server and mobile network services. Therefore, we need joint 4C for big data MEC is needed.
  • To deploy Edge AI, it is difficult to develop the distributed infrastructure or distributed model (e.g., federated learning scheme [2, 3]) to be able to learn on low-power and high-power edge devices with locally collected data.
  • To deploy Edge AI, it is difficult to find the best-suited domain-specific learning or prediction model such as finding object identification model among the various types of deep learning architectures Convolutional neural networks, Recurrent Neural Networks, etc.
  • To deploy Edge AI, it is difficult to tune parameters such as the number of layers (i.e., the depth of the network), types of layers (e.g., Convolutional, Recurrent and Fully Connected layers), and learning rate to improve the accuracy of the prediction model.

  • ▶ References

    1. H. Jin, L. Su, D. Chen, K. Nahrstedt, and J. Xu, “Quality of information aware incentive mechanisms for mobile crowd sensing systems,” in Proc. 16th ACM Int. Symp. Mobile Ad Hoc Netw. Comput., Jun. 22?25, 2015, pp. 167?176.
    2. E. Dave, “The internet of things: How the next evolution of the internet is changing everything.,” CISCO white paper 1, no. 2011, Apr. 2011, pp. 1?11. [Online]. Available: https://www.cisco.com/c/ dam/en_us/about/ac79/docs/innov/IoT_IBSG_0411FINAL.pdf
    3. E. Zeydan, E. Bastug, M. Bennis, M. A. Kader, I. A. Karatepe, A. S. Er, and M. Debbah, “Big data caching for networking: Moving from cloud to edge,” IEEE Commun. Mag., vol. 54, no. 9, pp. 36?42, Sep. 16, 2016.
    4. S. Ranadheera, S. Maghsudi, and E. Hossain, “Computation offloading and activation of mobile edge computing servers: A minority game,” IEEE Wireless Commun. Lett., vol. 7, no. 5, pp. 688?691, Oct. 2018.
    5. A. Ferdowsi, U. Challita, and W. Saad, “Deep learning for reliable mobile edge analytics in intelligent transportation systems,” CoRR, Dec. 2017. [Online]. Available: https://arxiv.org/pdf/ 1712.04135.pdf.
    6. Y. C. Hu, M. Patel, D. Sabella, N. Sprecher, and V. Young, “Mobile edge computing a key technology towards 5G,” ETSI White Paper, vol. 11, no. 11, pp. 1?16, 5 Sep. 2015.
    7. M. Patel, B. Naughton, C. Chan, N. Sprecher, S. Abeta, A. Neal, et al., “Mobile-edge computing introductory technical white paper,” White Paper, Mobile-edge Computing (MEC) Industry Initiative, Sep. 2014.
    8. O. Semiari, W. Saad, S. Valentin, M. Bennis, and H. V. Poor, “Context-aware small cell networks: How social metrics improve wireless resource allocation,” IEEE Trans. Wireless Commun., vol. 14, no. 11, pp. 5927?5940, Jul. 13, 2015.
    9. T. X. Tran, A. Hajisami, P. Pandey, and D. Pompili, “Collaborative mobile edge computing in 5G networks: New paradigms, scenarios, and challenges,” IEEE Commun. Mag., vol. 55, no. 4, pp. 54?61, Apr. 14, 2017.
    10. A. Ndikumana, S. Ullah, T. LeAnh, N. H. Tran, and C. S. Hong, “Collaborative cache allocation and computation offloading in mobile edge computing,” in Proc. 19th IEEE Asia-Pacific Netw. Operations Manag. Symp., Sep. 27?29, 2017, pp. 366?369.

    ▶ Achievements

    1. Ki Tae Kim, Yu Min Park, Choong Seon Hong, "Machine Learning Based Edge-Assisted UAV Computation Offloading for Data Analyzing," The International Conference on Information Networking (ICOIN 2020), January 7-10, 2020, Barcelona, Spain
    2. Nguyen H. Tran, Wei Bao, Albert Zomaya , Minh N.H. Nguyen and Choong Seon Hong, “Federated Learning over Wireless Networks: Optimization Model Design and Analysis,” IEEE International Conference on Computer Communications (INFOCOM 2019), April 29 - May 2, 2019, Paris, France
    3. Anselme Ndikumana, Choong Seon Hong, "Demand and Location-Based Multi-Access Edge Computing Deployment", 2018년 한국컴퓨터종합학술대회(KSC 2018), 2018.12.19~12.21
    4. Kyi Thar, Nguyen H. Tran, Thant Zin Oo, Choong Seon Hong, "DeepMEC: Mobile Edge Caching Using Deep Learning," IEEE Access, Vol.6, Issue 1, pp.78260-78275, December 2018
    5. Kyi Thar, Thant Zin Oo, Yan Kyan Tun, Do Hyeon Kim, Ki Tae Kim and Choong Seon Hong, "A Deep Learning Model Generation Framework for Virtualized Multi-access Edge Cache Management," IEEE Access