979-8-3503-6491-0/24/$31.00 ©2024 IEEE Conceptual Framework for the Optimization of Edge Device Workload Allocation in Wireless Body Area Networks Sachinthani Alahakoon Faculty of Information Technology, City University Malaysia, Petaling Jaya 46100, Malaysia sachi@gwu.ac.lk Mustafa Muwafak Alobaedy Faculty of Information Technology, City University Malaysia, Petaling Jaya 46100, Malaysia mustafa.theab@city.edu.my Yousef Fazea Department of Computer Sciences and Electrical Engineering, Marshall University, One John Marshall Drive, Huntington, WV 25755, USA, fazeaalnades@marshall.edu Abstract— Wireless body area networks (WBANs), a relatively new technology that has emerged in response to the exponential growth in the demand for healthcare, have shown themselves to be promising and are already being utilized extensively in the field of human health monitoring. Within the scope of this paper, a conceptual framework of an optimal workload allocation algorithm for edge computing-based WBAN systems is presented. Users can enhance their WBAN system by optimizing the workload allocation algorithms, which will result in a reduction in latency, energy consumption, job failure, and communication load. A final evaluation of the effectiveness of the proposed strategy was carried out, and the results of the experiment demonstrate that the proposed approach was successful in improving the quality of the system by lowering the number of transmission failures and the amount of power that was consumed. Keywords— Workload allocation, wireless body area networks, edge computing I. INTRODUCTION Through the enhancement of patient care, as well as the optimization of clinic operations and financial measures, the Internet of Medical Things (IoMT) contributes to the overall betterment of the healthcare industry [1]. Some important applications of IoMT include telemedicine services, with secure and reliable connectivity, healthcare professionals can remotely examine patients, provide consultations, and monitor their progress. IoMT streamlines healthcare operations by automating administrative tasks. This leads to cost savings, increased productivity, and improved overall efficiency in healthcare delivery. Furthermore, IoMT enables continuous remote patient monitoring and transmitting patient data to healthcare providers [1]. Patients' medical data collected via wearable healthcare devices such as wearable sensors and smart implants, enable continuous monitoring of chronic conditions like diabetes, and cardiovascular diseases also helps healthcare providers to identify trends, predict disease, and provide personalized treatments. This form of network is known as Wireless Body Area Network (WBAN) and comprises gateway devices, sensors, and additional devices located throughout the body [2]. Edge computing is gaining popularity in WBAN because it enables organizations to achieve near-real-time results by processing data as close to the source as possible. Its applications go beyond the medical field to include sports, entertainment, aerospace, and the military, among others. As a result, it adds greatly to both economic and social value [3]. A. WBAN for Health: Sensors and Computing A WBAN is a network that interconnects sensor nodes, enabling communication between individuals and objects. The utilization of a singular WBAN enables the monitoring variety of biomedical sensors that track a patient's vital signs. Biomedical sensors can monitor several vital indicators such as temperature, cardiac rate, breathing rate, electroencephalogram, electrocardiogram, blood pressure, and glucose level [4]. Every set of vital signs is represented by a unique type of clinical data that is completely different in format and content from the other data sets. As a result, the data collected by biomedical sensors are heterogeneous depending on the category of data, it requires different types of processing by medical professionals [5],[6]. WBAN can be divided into three types depending on the compute location. These are Cloud computing, fog computing, and edge computing-based WBAN systems. Edge computing is a computationally efficient, safe, private, and cost-effective approach for leveraging the Internet of Things (IoT) at scale. It eliminates the possibility of data breaches or network overloads, making it a superior alternative to fog computing and cloud computing. In addition, edge computing provides resilience and redundancy for mission-critical tasks. However, it does encounter challenges in the WBAN application process. The WBAN method utilizes several inbody, on-body, and off-body biomedical sensors, requiring the edge computing device to process large quantities of data simultaneously [3]. Due to the time-sensitive nature of such applications, it is necessary to identify the most desired processing requirements and prioritize them. Following the development of diverse WBAN models, researchers focus on enhancing the quality of service by addressing and overcoming technological challenges. With the help of rapid growth in biomedical sensors, low-power Integrated Circuits (ICs), and improvements in communication speed WBAN technology has emerged as an innovative, cost-effective solution for continuous health monitoring methods offering real-time updates. Consequently, the field of WBAN has attracted many researchers, particularly in optimizing resource and workload allocation strategies. These efforts aim not only to elevate service quality but also to bolster security measures for protecting sensitive medical data and to drive down the operational costs associated with health monitoring systems [3]. B. Challenges in Workload Optimization Prior to making critical medical decisions, collected raw data should be processed and analyzed. Therefore, the 2024 International Symposium on Networks, Computers and Communications (ISNCC) | 979-8-3503-6491-0/24/$31.00 ©2024 IEEE | DOI: 10.1109/ISNCC62547.2024.10759005 Authorized licensed use limited to: UNIVERSITY TEKNOLOGI MALAYSIA. Downloaded on February 07,2025 at 23:13:18 UTC from IEEE Xplore. Restrictions apply. processing of data is the most important aspect of WBANs. WBANs employ varied utilize different data processing strategies. Particularly, data processing for edge-based applications can be performed by edge processors themselves. Due to the urgent nature of many medical applications of WBANs, the most critical processing requirements must be identified. Processing requirements and processing priorities can be diverse significantly, reflecting the unique needs of each user. The WBAN system faces a significant challenge in finding an optimal strategy to accomplish the user's goal by allocating workload to edge nodes [1]. By using a suitable algorithm, it can achieve the desired performance, reduce processing delay or latency, lower power consumption, and enhance the throughput and growth of real-time applications that require local processing and storage capabilities [1]. The goal of this research is to introduce a conceptual framework to reduce processing delays and achieve the desired performance. Given the nature of the application, the most critical factors are speed, reliability, latency, and energy consumption. However, current routing algorithms cannot fully address these issues due to high propagation loss and challenging channel conditions [7]. The objective of this study is to significantly enhance the efficiency and speed of WBAN while reducing latency, lowering energy consumption, and minimizing the occurrence of task failures, by optimizing workload allocation algorithms for different WBANs with various edge device processing elements. Remote patient monitoring has gained significant popularity in recent years driven by population growth and the impact of the pandemic. Many hospitals and doctors have adopted the WBAN system for patient monitoring. Integrating edge computing with WBAN enables the health sector to extend its medical resources flexibly across locations while reducing time and costs. Edge computing has the ability to decrease latency, ultimately enhancing performance for the client [8]. Medical professionals increasingly use biomedical sensors to monitor various human body variations, aiming to maximize the benefits of resources within the WBAN system [9]. Considering the system integrates numerous sensors, the edge processor faces a significant workload. The processor in the edge device can discontinue or become overwhelmed, leading to slowdown and reduced efficiency coupled with high power consumption due to its heavy workload. Therefore, there must be an optimal workload allocation method for the WBAN system to achieve effective results. While there are various workload allocation methods for the edge device across multiple applications, these methods often fall short for time-critical applications. Hence, there is a pressing need to develop an optimal workload allocation algorithm specifically designed for time-critical edge computing applications such as WBAN. By employing optimizing workload allocation algorithms for specific applications with different edge processors, users can experience a significantly improved and more responsive WBAN system. II. LITERATURE SEARCH AND REVIEW Prior research has suggested multiple strategies for distributing edge workloads among distinct applications. The majority of research is centered around work offloading methods and resource allocation algorithms in order to optimize device time delay and energy usage in edge settings. The years between 2015 and 2020 witnessed a substantial surge in edge computing, which garnered considerable interest from both industry and academia. In [10], provides a comprehensive analysis of workload and resource allocation in edge-based WBANs. The assessment suggests that although there have been some investigations on workload distribution for fog and cloud computing, only the studies on task offloading and resource allocation were directly relevant to edge computing. Here are the pertinent studies that have been undertaken in recent years. In their work, Yuan et al., [7] presented a Two-Stage Potential Game-based Computation Offloading Strategy. This strategy attempts to optimize resource consumption in WBANs by taking into account task and user priorities. They initiate the process by articulating the challenge of maximizing system utility, specifically focusing on the quality of service (QoS) of activities. Computation offloading was modeled using reward, cost, and penalty functions. They enhance the feasibility of the algorithm and decrease the expense of computation by implementing a two-stage optimization approach. This method effectively resolves the problem of conflicting strategies in the strategy space of the possible game model [7]. The control mechanism presented in [11] utilizes bargaining game theory to facilitate resource sharing across the multi-access edge computing-assisted WBAN platform, which has limited computing and communication resources. The proposed approach evaluated the advantages of both intra- and inter-WBAN interactions. In 2023, Li et al. [12] discussed resource allocation and data offloading strategy for an edge-based telemedicine system. Utilizing cooperative game-based development, they allocated slots in WBAN to minimize system costs. Additionally, they employed a bilateral matching game-based system to optimize the data offloading problem in edge computing networks. Furthermore, researchers have recorded the practice of delegating tasks as a means of distributing workload. Zhang and Zhou introduced computational task offloading strategies that leverage mobile cloud computing and mobile edge computing for WBAN [13]. They devised a three-tier system along with an optimization technique to decrease expenses associated with energy usage and latency. A study was conducted to examine the efficacy of optimization methods for WBANs. Every optimization approach possesses unique advantages and disadvantages, simplifying the process of choosing the appropriate strategy by consulting comparative results [14]. Specific methodologies can be utilized to tackle problems in situations where conventional optimization approaches may not be appropriate, such as cases requiring discontinuous, nondifferentiable, stochastic, or very non-linear objective functions. In the next section, we will introduce a conceptual framework for workload allocation, aiming to build upon the foundational insights gleaned from the literature review. This framework will encapsulate the cutting-edge strategies and methodologies identified in recent research, tailored specifically for optimizing workload allocation in WBANs Authorized licensed use limited to: UNIVERSITY TEKNOLOGI MALAYSIA. Downloaded on February 07,2025 at 23:13:18 UTC from IEEE Xplore. Restrictions apply. and similar edge computing applications. Our focus will be on addressing the challenges highlighted in previous studies and proposing innovative solutions to enhance system performance, efficiency, and user satisfaction. III. A TRI-PHASE APPROACH TO ENHANCING WBAN EFFICIENCY The proposed conceptual framework consists of three phases to examine optimal workload allocation in edge-based WBANs. These are designing the WBAN model, developing the workload allocation algorithm, and carrying out experimental evaluations. Fig. 1. Fig 1: Overview of Methodology A. Phase 1 As a first phase of research, simulate different WBANs with a combination of different biomedical sensors using a suitable network simulation tool. Real-world deployments of WBANs can be expensive due to hardware, recruitment, and ethical considerations involving human subjects. Therefore, the phase 1 objective is designing WBANs to reflect realworld scenarios with different sensor sets and collect data on sensor node transmission times to the edge processor through simulations. To simulate WBAN models, various simulation tools such as PyLayer, Castalia, MoBAN, QualNet, Matlab, NS2, NS3, OMNET, OMNET++, OPNET, etc can be used. This research was chosen to be conducted with OMNET++ with the simple programming language Python, due to its user-friendly graphical user interface. Further, this tool supports modeling any network, such as Wireless sensor network (WSN), Software-Defined Networking (SDN), Ad-hoc, or Cellular (4G, 5G, and Beyond5G (B5G) ), can be used with both Ubuntu and Windows operating systems, and, most importantly, can be used with simple programming languages such as Python and C++ [15]. Different WBANs with different sets of biomedical sensors would then be designed to simulate real-world WBANs. Data transmitting time of each sensor node to the edge processor can be collected using simulation design. The characteristics of the proposed workload allocation method will be evaluated with the different WBANs with a combination of different biomedical sensors. Fig. 2 shows the WBAN simulation design with the OMNET++ software. Fig. 2. Fig 2: Simulating WBAN using OMNET++ Additionally, an online real-world data set will be used in the simulation. datasets should be found from the original sources or verify the data set validation with the original sources. The majority of government institutes, agencies, health departments, non-profit organizations, and educational or research centers offer their data sets available for academic use. In this research use electroencephalogram (EEG), electrocardiogram (ECG), Blood pressure, Temperature, and Oxygen saturation (SpO2) datasets for the simulation. But with the different WBAN models, we can use different sets of sensor data. These data sets processing utilizes a selected edge processor and determines the data processing time required for the algorithm's fitness function. This processing time is different for different edge processors. B. Phase 2 The second phase of this conceptual framework is to develop optimal workload allocation algorithms to enhance the system's quality. After successfully designing a WBAN, different workload allocation algorithms would be applied to find the optimum workload allocation algorithm for edgebased WBAN. Following a comprehensive review of the relevant literature, the PSO method was selected to optimize workload allocation in WBAN. Mainly considered their time complexity (
(),  − number of iterations,  − size of the given data set,  −dimension) and the performance. The fitness function was designed for the PSO model by considering the priority list of sensors. In the fitness function,   added to add priority value to the sensors according to the priority list. These priority values can be changed according to the user requirements. With a few changes, this fitness function can be used for different applications.                Where n = number of sensors,      =  th sensor’s processing time +  th sensor’s data transferring time,    =  th sensor scanning count for collecting data,   = weight of  th sensor Authorized licensed use limited to: UNIVERSITY TEKNOLOGI MALAYSIA. Downloaded on February 07,2025 at 23:13:18 UTC from IEEE Xplore. Restrictions apply. C. Phase 3 Phase three of the conceptual framework is carrying out experimental evaluations. The objective of the third phase is to evaluate the performance of the workload allocation algorithm across various WBAN configurations. Experiments would be carried out to determine the optimum workload allocation method that is independent of the WBAN for each processing element. There are several edge computing processors that can be used with WBANs, such as NVIDIA Jetson Nano, Arduino Yun, Banana Pi M3, Raspberry Pi, and ESP 32 [16]. In this part of the research, at least two different processing elements (Raspberry Pi4 and ESP 32) will be used. The Raspberry Pi4 and the ESP32 are two of the most popular microcontrollers on the market that can be used as edge processors. They are two small, lowpower consumption, low-cost microcontrollers with high community support. Both processors can interface with other systems to provide Wi-Fi and Bluetooth functionality through their SPI / SDIO or I2C / UART interfaces [17]. The ESP32 however has a faster processor and more flash memory which results in more power draw. Fig. 3 depicts the fundamental experimental model utilized in the research. Start by evaluating the network quality of the system, focusing on the communication strength between the edge node and cloud server. This process involves sending data packets to the cloud and measuring the time it takes for those packets to be received. Fig. 3. Fig 3: Basic experiment model Then update the data sending frequency of the algorithm (frequency of data sending to the cloud from the edge node). Next, execute the developed PSO algorithm on the edge processor in order to optimize the allocation of workload. Once the algorithm is implemented, the number of sensor scans per transmission is updated to get the maximum benefit from the system. Then the processed data is sent to the cloud and end-user for storing and analyzing information that can be utilized for disease monitoring, diagnosis, and therapy. Next, evaluate the output and compare the efficiency of the experimented algorithm with other algorithms or without any optimization techniques. To accomplish this, power consumption, and transmission failures of the system will be measured, as illustrated in Fig. 3. This conceptual framework and algorithm will optimize workload allocation in edge processors by considering the network quality. Therefore, with the developed conceptual framework, it would be possible to update data sending frequency according to the communication strength and scanning sensors maximum time and collect data. Due to the optimum workload allocation, system data transmission failures and power consumption will be reduced. IV. E XPERIMENTAL RESULTS USING CONCEPTUAL FRAMEWORKCONCLUSION The proposed conceptual framework was tested and validated through an experiment. For this research, we selected the PSO method to optimize workload allocation in edge-based WBAN after conducting a thorough review of the relevant literature. The Raspberry Pi4 microcontroller serves as an edge processor in this experiment. In this section, we discussed the experimental results of the present conceptual framework for optimal workload allocation in edge-based WBANs. This distribution has sample size 30, which is generally considered large enough for the CLT to apply. For the power consumption of PSO at 25-line quality value, data set indicated a 95% confidence interval of 3.562 ± 0.0549. This interval suggests that we can be 95% confident that the true mean performance of the PSO algorithm lies within the range of 3.5071 to 3.6169, demonstrating the algorithm's consistent performance across the dataset. As shown in Fig. 4, the experimental results are based on the number of unsuccessful transmission attempts versus network line quality. The experiment was carried out for both workload allocation using the PSO algorithm and without any optimization technique. The graphic shows that compared to not using any optimization strategy, there are fewer unsuccessful transmission attempts when the PSO algorithm is applied. Fig. 4. Fig 4: Unsuccessful data transmission attempts on each line quality In both scenarios, the number of unsuccessful transmission attempts increases as line quality decreases. However, the graph shows that the number of unsuccessful transmission attempts of workload allocation without any optimization technique is almost twice the amount of unsuccessful transmission attempts of workload allocation with PSO. This is because workload allocation with PSO helps to reduce unsuccessful transmission attempts due to the adaptation of network line quality. Based on the line quality, the PSO adjusts the sending frequency and the sensor scanning frequency. Authorized licensed use limited to: UNIVERSITY TEKNOLOGI MALAYSIA. Downloaded on February 07,2025 at 23:13:18 UTC from IEEE Xplore. Restrictions apply. Fig. 5. Fig 5: power consumption on each line quality Fig. 5 presents the power consumption of the system for different network quality with the PSO algorithm and without any optimization technique. As observed from the figure there is a higher power consumption for workload allocation without any optimization technique. Since Raspberry pi4 uses 3.5 W to operate, there is an average 60% increase in extra power consumption for workload allocation without any optimization techniques than workload allocation using PSO. Experimental results show that the employment of the PSO algorithm enhances the quality of a system by reducing unsuccessful data transmission and power consumption. These findings pave the way for more resilient and energyefficient WBAN systems, with significant potential to improve patient care and operational efficiency in healthcare applications. V. CONCLUSION In this study, we have presented a conceptual framework for optimizing workload allocation in edge device-based WBAN. This research contributes to the body of knowledge in the WBAN field by providing the optimal workload allocation algorithm and experimental framework. It helps in conducting experiments which are having high costs in realworld scenarios. The proposed method effectively increases the WBAN system performance by reducing data transmission failures and power consumption. In the future, this framework will be used to experiment with different optimization techniques and different edge processors. Also, we intend to work toward a proactive workload allocation algorithm that enhances the quality of the WBAN system. Researchers may use this idea to develop new optimal workload allocation methods to further improve the efficiency of edge-based WBAN systems. REFERENCES [1] G. Aloi, G. Fortino, R. Gravina, P. Pace, and G. Caliciuri, “Edge Computing-Enabled Body Area Networks,” in 2018 32nd International Conference on Advanced Information Networking and Applications Workshops (WAINA), IEEE, May 2018, pp. 349–353. doi: 10.1109/WAINA.2018.00110. [2] Y. Qu, G. Zheng, H. Ma, X. Wang, B. Ji, and H. Wu, “A Survey of Routing Protocols in WBAN for Healthcare Applications,” Sensors 2019, Vol. 19, Page 1638, vol. 19, no. 7, p. 1638, Apr. 2019, doi: 10.3390/S19071638. [3] L. Zhong et al., “Technological Requirements and Challenges in Wireless Body Area Networks for Health Monitoring: A Comprehensive Survey,” Sensors, vol. 22, no. 9, p. 3539, May 2022, doi: 10.3390/s22093539. [4] J. Gonzalez, J. Hunt, M. Thomas, R. Anderson, and U. Mangla, “Edge computing architecture and use cases,” IBM developer, Apr. 2022, [Online]. Available: https://developer.ibm.com/articles/edgecomputing-architecture-and-use-cases/ [5] K. Karunanithy and B. Velusamy, “Edge device based efficient data collection in smart health monitoring system using wireless body area network,” Biomedical Signal Processing and Control, vol. 72, p. 103280, Feb. 2022, doi: 10.1016/j.bspc.2021.103280. [6] L. R and V. P, “Wireless Body Area Network (WBAN)-Based Telemedicine for Emergency Care,” Sensors, vol. 20, no. 7, p. 2153, Apr. 2020, doi: 10.3390/s20072153. [7] X. Yuan, H. Tian, H. Wang, H. Su, J. Liu, and A. Taherkordi, “Edge- Enabled WBANs for Efficient QoS Provisioning Healthcare Monitoring: A Two-Stage Potential Game-Based Computation Offloading Strategy,” IEEE Access, vol. 8, pp. 92718–92730, 2020, doi: 10.1109/ACCESS.2020.2992639. [8] K. Sadatdiynov, L. Cui, L. Zhang, J. Z. Huang, S. Salloum, and M. S. Mahmud, “A review of optimization methods for computation offloading in edge computing networks,” Digital Communications and Networks, Mar. 2022, doi: 10.1016/j.dcan.2022.03.003. [9] M. T. Arefin, M. H. Ali, and A. K. M. F. Haque, “Wireless Body Area Network: An Overview and Various Applications,” Journal of Computer and Communications, vol. 05, no. 07, pp. 53–64, 2017, doi: 10.4236/jcc.2017.57006. [10] S. Alahakoon and R. Logeswaran, “Review on Workload and Resource Allocation in Edge-Based Wireless Body Area Networks,” in 2023 IEEE 13th Symposium on Computer Applications & Industrial Electronics (ISCAIE), IEEE, May 2023, pp. 304–308. doi: 10.1109/ISCAIE57739.2023.10165335. [11] S. Kim, “A New Two-Stage Bargaining Game Approach for Intra- and Inter-WBAN Management,” Mobile Information Systems, vol. 2021, pp. 1–10, Nov. 2021, doi: 10.1155/2021/5798741. [12] Y. Li, Y. Wang, S. Chen, X. Huang, and T. Huang, “Resource Allocation and Data Offloading Strategy for Edge-Computing-Assisted Intelligent Telemedicine System,” vol. 23, no. 10, May 2023, doi: 10.3390/s23104943. [13] R. Zhang and C. Zhou, “A Computation Task Offloading Scheme based on Mobile-Cloud and Edge Computing for WBANs,” in ICC 2022 - IEEE International Conference on Communications, IEEE, May 2022, pp. 4504–4509. doi: 10.1109/ICC45855.2022.9838921. [14] N. Bilandi, H. K. Verma, and R. Dhir, “Performance and evaluation of energy optimization techniques for wireless body area networks,” Beni- Suef University Journal of Basic and Applied Sciences, vol. 9, no. 1, p. 38, Dec. 2020, doi: 10.1186/s43088-020-00064-w. [15] Network Simulation Tools, “Network Simulation Tools Projects for Research Scholars,” Network Simulation Tools. [Online]. Available: https://networksimulationtools.com/network-simulation-tools-projects/ [16] C. Nicholson, “5 processor architectures making machine learning a reality for edge computing. Enable,” Enable Architect. [Online]. Available: https://www.redhat.com/architect/processor-architecturesedge-computing [17] A. Fanariotis, T. Orphanoudakis, K. Kotrotsios, V. Fotopoulos, G. Keramidas, and P. Karkazis, “Power Efficient Machine Learning Models Deployment on Edge IoT Devices,” Sensors, vol. 23, no. 3, p. 1595, 2023, doi: https://doi.org/10.3390/s23031595. Authorized licensed use limited to: UNIVERSITY TEKNOLOGI MALAYSIA. Downloaded on February 07,2025 at 23:13:18 UTC from IEEE Xplore. Restrictions apply.