Multi-Access Edge Computing Offloading Methods for Maritime Monitoring Sensor Networks

Click the blue text above to follow us!

Multi-Access Edge Computing Offloading Methods for Maritime Monitoring Sensor Networks

Su Xin¹, Wang Ziyi¹, Wang Yupeng², Zhou Siyuan³

1 College of Internet of Things Engineering, Hohai University, Changzhou, Jiangsu 213002, China

2 School of Electronic Information Engineering, Shenyang Aerospace University, Shenyang, Liaoning 110136, China

3 School of Computer and Information, Hohai University, Nanjing, Jiangsu 211106, China

Abstract:

Multi-access edge computing (MAC) can effectively ensure low latency and high reliability in data transmission for maritime monitoring sensor networks (referred to as sensor networks) and various related maritime applications. In coastal scenarios, two offloading models are established: multi-user single-hop unicast (MSU) and multi-user multi-hop unicast (MMU), based on the distribution of edge computing resources. By separating the optimization objectives using mixed-integer nonlinear programming, transmission power is effectively allocated, and offloading decisions are made using an improved artificial fish swarm algorithm (AFSA). The results show that the proposed optimization algorithm can reduce network latency by nearly 19% compared to traditional solutions. In offshore scenarios, a distant sea MSU offloading model is established, and a reasonable channel allocation algorithm is proposed based on network connectivity probability. The results indicate that when network connectivity time is sufficient, the number of sub-channels allowed for allocation can be increased, reducing network latency; when network connectivity time is limited, the number of maritime user devices to be offloaded can be controlled to ensure network transmission latency.

Keywords: Maritime monitoring sensor networks; Multi-access edge computing; Artificial fish swarm algorithm; Channel allocation

Citation format:

Su X, Wang Z Y, Wang Y P, et al. Multi-access edge computing offloading in maritime monitoring sensor networks[J]. Chinese Journal on Internet of Things, 2021, 5(1): 36-52.

SU X, WANG Z Y, WANG Y P, et al. Multi-access edge computing offloading in maritime monitoring sensor networks[J]. Chinese Journal on Internet of Things, 2021, 5(1): 36-52.

1 Introduction

The maritime monitoring sensor network (referred to as sensor network) is an important part of the future smart marine information network, providing various monitoring applications and serving as a vital platform for gathering various data on marine space, environment, and resources. The implementation of all-weather, fully automated marine observation, situational awareness, marine information transmission, and comprehensive maritime service poses demands for real-time positioning, emergency rescue, and low-latency information services, particularly in military applications. These demands place high requirements on the maritime monitoring sensor network to support low-latency and high-reliability monitoring applications. The multi-access edge computing (MAC) technology, aimed at the next generation of mobile networks, can effectively support these maritime business and application needs by “downgrading” cloud data centers to the network edge, enabling them to provide computing, storage, and communication capabilities closer to users. However, the new maritime monitoring sensor network system based on MAC still faces several scientific challenges.

Complex maritime monitoring applications can lead to overloaded data processing in local network areas, with a sharp increase in network overhead in certain regions. Achieving flexible adaptation of resources in maritime monitoring sensor networks is the primary problem to be solved.

Compared to traditional terrestrial cellular networks and vehicular networks, the environmental factors of maritime monitoring sensor networks are complex, with significant node differentiation. In particular, in coastal scenarios, the node density of maritime monitoring sensor networks is relatively high, with uneven distribution of edge computing capabilities and complex resource scheduling; in offshore scenarios, the node density is lower, and network connectivity is easily affected by weather and harsh sea conditions, making service continuity difficult to guarantee. Proposing edge data offloading models and algorithms that adapt to different maritime scenarios is also an issue that needs to be addressed.

Based on research related to MAC offloading technology and the challenges faced by maritime monitoring sensor networks, this paper aims to meet the demands for low latency and high reliability in maritime applications and conducts research on MAC offloading technology in maritime monitoring sensor networks. The research results and main contributions are as follows.

This paper establishes two different multi-user offloading models for coastal and offshore scenarios in maritime monitoring sensor networks. For coastal mobile nodes with sufficient and limited computing resources, MSU and MMU offloading models are established respectively. Corresponding priority functions are set to address issues of multi-user contention for channel resources under the MSU offloading model and significant differences in computing resources of first-level offloading nodes under the MMU offloading model. Meanwhile, for offshore scenarios with limited mobile nodes, computing resources, and network connectivity time, a distant sea MSU offloading model is established. To alleviate data offloading blockage, orthogonal frequency division multiple access (OFDMA) technology can be used to allocate physical channels.

In coastal scenarios, the proposed mixed-integer nonlinear programming optimization problem is separated into two sub-problems, effectively allocating transmission power through mathematical optimization, and offloading decisions are effectively made using an improved traditional artificial fish swarm algorithm. In offshore scenarios, a low-complexity sub-channel allocation algorithm for distant sea MSU offloading is proposed to solve the optimal matching problem between multi-users and sub-channels under sufficient and limited network communication time.

2 Related Work

MAC is a core technology for future mobile communications. Combining the advantages of this technology with the characteristics of maritime network edge computing, this paper proposes an optimized offloading scheme for different maritime scenarios. However, the current achievements of applying MAC to maritime monitoring sensor networks to improve network performance are relatively scarce. Therefore, this paper focuses on analyzing and discussing research results related to multi-access edge computing offloading schemes based on terrestrial scenarios, which provide important references for the innovations presented in this paper.

2.1 Single User Multi-Access Edge Computing

The single-user multi-access edge computing model can adapt well to the limitations of mobile terminals, leading to a better user experience by improving the utilization of computing and storage resources of mobile terminals. Literature [8] offloads all terminal data to the edge server, proposing a random offline strategy based on pre-computation to optimize radio computing resource scheduling and offloading. The application of multi-access edge computing can not only save energy consumption but also significantly reduce latency. Literature proposes an adaptive optimization-based computing offloading method and uses an adaptive algorithm based on Lyapunov theory to optimize the weighted sum of latency and energy consumption. Although complete offloading schemes have improved latency and energy consumption compared to local processing, the aforementioned schemes treat terminal data as a whole, only achieving either complete local processing or complete offloading, limiting the overall efficiency of the system.

Data splitting offloading schemes can take two forms: one divides data into multiple subtasks, each deciding whether to offload; the other introduces variables to split data into two parts based on whether they can be offloaded. Literature considers the locally processed part of data that cannot be offloaded while averaging the offloadable part into a certain number of subtasks, achieving maximum energy difference between local computation and offloading through a low-complexity suboptimal algorithm to obtain optimal offloading decisions for each subtask. However, the above research results focus on energy-sensitive user demands, and the average splitting of data lacks practical significance. Therefore, to meet diverse user needs, literature proposes a weighted sum optimization problem for multi-task latency and energy consumption, employing alternating solutions and suboptimal solution methods for resource allocation and offloading decisions. Literature explores the application of dynamic voltage scaling technology in multi-access edge computing, introducing variable segmentation for pending data. It also proposes schemes to reduce energy consumption and shorten latency to meet different user demands. Facing the diverse maritime applications, the single-user model is not suitable for the development trend of future smart oceans. The increasingly complex data migration environment necessitates research on resource allocation issues under multi-user models to further enhance user experience.

2.2 Multi-User Multi-Access Edge Computing

Literature proposes a power minimization problem under task buffer stability constraints and optimizes resource allocation for users in each time slot through the Lyapunov algorithm. Literature applies time-division multiple access technology to multi-access edge computing offloading models, proposing a joint communication and computing resource allocation optimization algorithm to reduce latency. Based on literature, literature utilizes OFDMA technology to propose an energy-saving computing offloading and resource allocation algorithm to optimize system overhead. Literature combines the advantages of time-division multiple access and OFDMA technologies to establish a multi-user offloading model, analyzing the energy optimization problem under limited computing resources and proposing suboptimal algorithms for efficient resource allocation under both models. Literature only considers data uplink transmission latency, while literature comprehensively considers both uplink and downlink transmission latency, hierarchically solving the latency optimization problem, optimizing inner-layer resource allocation using the bisection method, and optimizing outer-layer offloading decisions using a greedy heuristic algorithm. However, the unloading model of this method ignores the computation latency of edge nodes (only considering offloading transmission latency) and the offloading energy consumption constraints, failing to meet the practical application needs. Based on the above analysis, this paper applies two data partitioning methods under the single-user model to the multi-user model, establishing a more flexible partial offloading model. Furthermore, utilizing OFDMA technology to partition channels further achieves low-latency and high-reliability transmission for maritime monitoring sensor networks. Some maritime nodes (communication and computing) have limited resources; therefore, the offloading model designed in this paper considers the computation latency and transmission energy consumption constraints of nodes, making it more aligned with the characteristics of maritime monitoring sensor networks.

3 Edge Computing Offloading in Maritime Monitoring Sensor Networks

3.1 Architecture of Maritime Monitoring Sensor Networks

The edge computing offloading model of maritime monitoring sensor networks is shown in Figure 1, primarily composed of oceanic user equipment (OUE) and oceanic edge computing nodes (OECN).

Multi-Access Edge Computing Offloading Methods for Maritime Monitoring Sensor Networks

Figure 1 Edge computing offloading model of maritime monitoring sensor networks

OUE has limited local computing capabilities and signal coverage, is energy-sensitive, and can conduct maritime monitoring tasks in an all-weather and fully automated manner, specifically including lightweight unmanned submersibles, marine buoys, small vessels, and unmanned ships. Lightweight unmanned submersibles can be used for port reconnaissance and surveillance, mine detection, and underwater surveys of specific destinations. Marine buoys can continuously provide required oceanographic, hydrological, and meteorological data for marine scientific research, offshore oil (gas) development, and port construction. Small vessels have sensing and communication capabilities and can perform tasks related to navigation surveying, marine environment monitoring, and marine resource development. Unmanned ships can replace divers for operations such as shipwreck salvage, deep-sea exploration, and underwater cable laying.

OECN has sufficient internal energy, strong local computing capabilities, and extended signal coverage, effectively conducting related maritime monitoring tasks, specifically including coastal base stations (SBS) and medium to large vessels. Medium to large vessels can serve as edge computing nodes, processing outsourced tasks from OUE nodes in real-time, and can also act as relay nodes to achieve task forwarding, connecting to surface self-organizing networks to expand the coverage of coastal base station networks. Coastal base stations serve as aerial interfaces, connecting maritime monitoring equipment terminals to land cloud servers, and can also act as edge servers to effectively support high-reliability and low-latency related services.

3.2 Edge Computing Offloading in Coastal Scenarios

Figure 1 describes the real-time parallel execution scenario of multiple tasks in coastal scenarios. At this time, the maritime network local area generates a large amount of data that needs to be processed in real-time, which can be fused and clustered and offloaded to nearby OECN for processing. If OUE is close to the coastal base station, it can directly offload tasks to the adjacent base station, referred to as first-level MSU offloading. If OUE is far from the coastal base station, various tasks can be offloaded to nearby medium to large vessels for real-time processing, or some tasks can be offloaded to the coastal base station for processing based on the real-time load of its own server, referred to as second-level MMU offloading.

The first-level MSU offloading scenario is suitable for situations where maritime network edge computing resources are sufficient, considering the single-hop offloading mechanism between K OUE and a single base station. Literature establishes a multi-user offloading model to a single MAC server system, utilizing OFDMA technology to divide the channel into N sub-channels n∈{1,2,…,N} to process a group of OUE offloading data. However, this model only considers complete offloading situations, where each user can only choose to process locally or completely offload, leading to low flexibility and latency efficiency.

As shown in Figure 1, each OUE computing task is divided into S subtasks of different sizes, and the subtask set of a single OUE is represented as Task_i∈{Task₁,Task₂,…,Task_S}. Each subtask can be processed locally or offloaded to nearby OECN for processing. To avoid resource competition and reduce conflict probability, a corresponding relationship between single subtask and sub-channel can be established.

The second-level MMU offloading scenario shown in Figure 1 is suitable for situations where maritime network edge computing resources are limited. Each subtask can choose to use medium to large vessels within its communication range as its first-level offloading node for data processing, or it can use that node as a relay node to transmit to a second-level offloading node (coastal base station) with sufficient computing resources for processing. Ultimately, the delays consumed by both offloading methods and local processing are compared to determine the optimal processing method for that task. Similar to the MSU offloading scenario, the MMU offloading scenario adopts a corresponding relationship between subtask and sub-channel for computing offloading, where the sub-channel set can be represented as…

Furthermore, computing resources are limited for first-level OECN such as medium to large vessels. Multiple OUE offloading their computing tasks to the same first-level OECN at the same time increases the waiting delay for different OUE subtasks at that node, reducing offloading benefits. Therefore, it can be assumed that in the MMU offloading scenario, there is also a corresponding relationship between OUE and first-level OECN (i.e., each OUE can only select a different mobile node from other OUE as its first-level OECN), only needing to consider the waiting delay of each subtask at its corresponding first-level OECN without needing to consider the waiting delay of multiple OUE subtasks at the same node, thereby improving the benefits of computing offloading.

In the second-level MMU offloading scenario, when selecting the optimal task offloading coastal base station, the connectivity probability between maritime network nodes must be considered. Literature derives the connectivity probability function P_s for OUE when it is outside the SBS signal coverage range, through relay nodes connecting to SBS. By grouping each pair of adjacent SBS, the connectivity probability function P_s between OUE and each group of SBS can be derived, and according to the principle of maximum connectivity probability, the optimal SBS can be selected, with the relevant algorithm pseudocode shown as Algorithm 1.

Algorithm 1 OUE Selection Strategy for SBS

Input: Number of selectable SBS, mobile ship density ρ_mob, distance between a group of SBS (SBS₁, SBS₂), SBS communication radius (same for two SBS in a group), distance between SBS₁ in a group.

Output: Maximum value of P_s and corresponding SBS.

Initialize

…

End

…

End

…

End

…

Contact Us:

Tel:010-81055683

010-81055483

010-81055691

E-mail:[email protected]

http://www.infocomm-journal.com/wlw

http://www.wlwxb.com.cn

The “Journal of Internet of Things” aims to publish innovative academic papers in the field of Internet of Things and related interdisciplinary research, showcasing China’s research achievements, reflecting the country’s cutting-edge research level, providing academic support for the rapidly developing IoT technology innovation and industry services, and promoting the development of IoT technology and applications in China.

Multi-Access Edge Computing Offloading Methods for Maritime Monitoring Sensor Networks

1 Introduction

2 Related Work

2.1 Single User Multi-Access Edge Computing

2.2 Multi-User Multi-Access Edge Computing

3 Edge Computing Offloading in Maritime Monitoring Sensor Networks

3.1 Architecture of Maritime Monitoring Sensor Networks

3.2 Edge Computing Offloading in Coastal Scenarios

Related posts

Leave a Comment Cancel reply