HiLSeR: Hierarchical learning-based sectionalised routing paradigm for pervasive communication and Resource efficiency in opportunistic IoT network

https://doi.org/10.1016/j.suscom.2021.100508Get rights and content

Highlights

  • Leveraging social aspect in opportunistic-IoT communication through pervasive intelligence to enable sustainable computing.

  • Enabling resource efficiency to minimize energy consumption, system overheads, buffer utilization and dropped messages.

  • Sectionalizes the network topology based on node characteristics and employ grouping in intelligent transmission.

  • Enables message routing using a combination of controlled flooding and opportunistic sector-based decentralized transmission.

  • Uses designated tier specific apex nodes for inter-cluster communication that further increases efficiency.

Abstract

Opportunism in the Internet of Things is the latest necessity arising in the IoT network communication development when we encompass a sustainable approach to routing. Such a requirement arises from the need for device-hop based networking to eliminate a dedicated end-to-end physical network for pervasive communication and promotes local data and computation sharing mechanisms as well as sustainably utilising the ubiquitous mobility aspect of such applications in the contemporary times. However, the functionality of opportunism runs into its own set of impasse; intermittent connectivity, limited resources, need for intelligence in routing, to name a few. Leveraging the ability of ML to tailor represent features to our advantage, this paper proposes a scheme to sectionalize the network topology based on node characteristics and employ grouping in intelligent transmission. The proposed scheme HiLSeR enables message routing using a combination of controlled-parameterized flooding and opportunistic sector-based decentralized transmission. Hierarchical learning, a multi-dimensional data conduct based soft clustering paradigm, is used for topology sectionalization and routing decision making. The performance of the proposed scheme is evaluated against contemporaneous RLPRoPH, GMMR, KNNR and Firefly PRoPHET protocols with ONE based simulations. The performance and sustainability performance is compared on various parameters such as Energy Unit per message, Dead node Percentage, Overhead Ratio, Average Latency and Success Ratio to show the enhanced performance. HiLSeR has an average successful delivery rate of 0.911 averaging out at 0.86775, in comparison to RLPRoPH, GMMR, KNNR, Firefly PRoPHET, the proposed scheme performs 12.85 %, 5.59 %, 61.29 %, 18.50 % and 88.33 % better respectively.

Introduction

Our collective history of communication reflects and echoes the changes in human behaviour as well as technologies. We currently are living in a hyper-connected, ubiquitous, nonlinear and chaotic era of communication [1]. This state of has been unprecedented from a historical perspective and extremely transient, each year we are making longer leaps which is transforming this ecosystem which is already extremely dynamic. Over the past few decades, pervasive communication has transformed from being reflective of the number of channels of communication but encompasses the diversity in those channels. Sustainability in computing has gained significant traction and prevalence in recent times with research communication [1]. Their application particularly in mobile intelligence has risen and has transformed it to a necessity with increasing IoT based devices. With communication being distributed dynamically across various devices, pervasive communication has given rise to a myriad of challenges including but not limited to integration heterogeneous devices, asymmetricity, intermittent connectivity, high error rate among others. This paves the way for the development of ubiquitous systems with mobile nodes which can collaborate opportunistically for facilitating inter-device data transfer [2].

With pervasive communication, “chaos” induced by the diversity of communicating devices or nodes gets injected in the system, but this also gives is “opportunity” to leverage pervasive paradigm to mitigate this. We look at this from the “social” aspect of computing and associate a node’s mobility to and find patterns using intelligent techniques in a ubiquitous manner to enable computing task. This forms the basis for the problem statement our paper deals with. In our paper, we are addressing the challenges in pervasive communication through the use of a context-aware intelligent technique for routing data among IoT nodes in an opportunistic manner.

From a conventional standpoint, the Internet of Things abbreviated as IoT aims to link every physical object in order to build and a cohesive infrastructure of IoT with a confluence of functionality that is available. Opportunistic networks are characterized by extended latency and unstable network topology where network contacts are intermittent and there exists an absence of a route between the source and goal node for most of the time. The following characteristics or a combination of these are often attributed to Oppnets which includes scant connectivity, persistent partitioning, immense delays, asymmetry in information transfer rates and less reliability. Further, the node configuration or topology is dynamic and an end to end node connectivity is absent. In order to bring this unhindered connection between devices, IoT deals with a large magnitude of information through networks that are inherently heterogeneous. Not only opportunistic networks along with their nodes that are in continuous motion makes this heterogeneous integration easy but lowers reliance on a central system based infrastructure. They catalyze the spread of infrastructure that is large in size and inherently heterogeneous. The devices in these networks have connectivity to the internet but this connectivity is not always there. Depending on the context and geographical challenges this may be restricted. Here the contacts are also known as opportunities are intermittent [3].

The link connectivity or performance is extremely wavering and an end to end connectivity is for a brief unpredictable time hence the conventional Internet protocol suite will not function in this scenario. The large delays in propagation and the irregular queuing may lead to breaking and this adds to the issue as the traditional protocols make an implicit assumption of receiving prompt acknowledgement hence they fail to operate in these scenarios. Additionally, there exists asymmetric data rates and a high error rate in transmission. One such solution to enhance the above impediment is to make use of the mobile nature of the nodes in the network and the ability to locally forward data. When opportunistic contacts are made so the intermediate nodes can forward the data successively and forward the data to the destination.

This leads us to the hybrid concept of Opportunistic IoT or OppIoT networks which essentially has IoT devices as the nodes with their individual properties adding to the specificity. In principle, OppIoT attributes similar properties of delay tolerance in comparison to traditional Oppnets. This is of great utility in several scenarios where the architecture of the network is lacking, key information pertaining to the configuration is absent and the nodes are following random mobility models. Conventional IoT enabled networks contains IoT devices and thorough variable contact opportunities message dissemination happens [4]. The forwarding occurs here in a similar manner to that of Oppnets when two or more nodes encounter within the communicable distance, this is followed by a delivery probability-based decision-making paradigm. Consider the encounter of two nodes, when this connection is made the buffer of the sender node containing the message has to make a decision pertaining to the transmission of the message. The decision pertaining to whether to forward the message or not is termed as the next-hop selection. Instinctively, the message should be forwarded by the sender to the candidate node in case the probability of, direct or indirect, delivery of the message by the candidate is large enough. If a message is forwarded frequently this can result in an excess of loss of packet and overloading of buffer overhead. Alternatively, if the forwarding is less frequent owing to too high thresholds, this would directly lead to plummeting numbers of delivered messages. Hence, the routing schemes follow a store-carry-forward paradigm to enable routing in this challenging scenario. It may be noted, OppIoT’s forwarding nature is similar to Oppnets and hence the same protocol designs can be applied to both the classes [5].

Another concept this paper explores is the Self-configuration under the ambit of network management. Under this, a node is able to self-configure, or in other words, is able to morph its parameters in an autonomous manner in alignment to the environment. This is in particular quite helpful for an Opp-IoT scenario where conditions are quite dynamic. Opportunistic IoT requires increased scalability and augmented performance which self-configuration can cater to. Although this paradigm in itself introduces a litany of challenges hence implementation would require careful calibration to avert complexities arising due to distribution, loop distribution and congestion due to traffic. Irrespective, self-configuration paradigm is quite useful to account for the dynamic behaviour and implement intelligent routing, something which our paper explores in detail. This emulates an extremely integrated relationship between smart things and their opportunistic connections with humans, this allows the formation of opportunistic community’s dependent on the movement that forms the social behaviour of the humans. OppIoT assists in increasing the coverage of sparse IoT enabled networks that have mobile nodes. To summarize, there are a variety of benefits in using opportunism in the IoT scenario. This factors in and eliminates IoT networks' dependency on a well-defined infrastructure which provides an alternate communication strategy such as but not limited to Wi-Fi, Bluetooth etc. The opportunistic contacts formed here factors in the inherent heterogeneity of the device. Additionally, it enables the network to cater to sparsely connected regions and address the bounds of geographical constraints.

The proposed scheme makes use of ML-based sectionalizing of the network topology based on node characteristics and employs grouping in intelligent transmission. The proposed scheme HiLSer enables message routing using a combination of controlled-parameterized flooding and opportunistic sector-based decentralized transmission. Hierarchical learning, a multi-dimensional data conduct based soft clustering paradigm, is used for topology sectionalizing and routing decision making.

Sustainability has been a key area of consideration in a myriad of domains, including the development of routing algorithms. Research literature highlights the importance of resource management and energy efficiency while observing. In an IoT network, the nodes are constrained on various resources such as constrained energy, buffer storage and processing power. The largest source of energy consumption for IoT nodes is transmission and forwarding of data over a long distance. Further, the leading cause of sharp decay in delivery probability in a network is the loss of power and a full buffer in contrast to hardware failures. This is in particular quite significant in cases such as opportunistic networks, as considered in the objective of this paper, as the connectivity is intermittent and the end to connectivity is absent. Hence it is of significant traction that routing algorithms developed for this scenario factor in resource management and make energy considerations. This paper proposes a pervasive and ubiquitous intelligence-based approach that ensures the reliability of transmission while being resource-efficient. The proposed scheme, HiLSeR, adopts a multi-objective multi-tier approach employing pervasive communication and ubiquitous context data to constrain the flooding and achieve high delivery reliability and resource efficiency. The ubiquitous context-aware approach limits the unnecessary flooding and a tiered clustering using intelligent optimisation the data packets are routed. The methodology of the approach is detailed in the latter section of this composition along with relevant comparative results to attest to this claim.

Fig. 1 shows the OppIoT architecture and networks. Several routing schemes have adopted different approaches to ascertain this probability and make routing considerations, the same has been explored in Section 2 of this paper. Every routing protocol is developed with the aim of minimizing the overhead and being energy efficient. In this scenario, developing a routing paradigm raises a myriad of proximate and underlying issues, primarily they are [6,7] Uncertainty in Transmission, Allocation of Resources, Performance, Reliance and Security. In order to understand the network scenario, we need to understand the social relations between the nodes. In theory, at proximate level nodes in an opportunistic network are supposed to follow a random mobility model. However, this is not true in a practical sense as nodes in any network are essentially devices that share some degree of social relationship among one and other. These nodes may be IoT devices and based on their operation bear characteristics that do not make their motion truly random [8]. Alternatively, these nodes may be carried by humans, animals or objects that share some pattern in their movement and possess a tendency to form sub-groups or clusters within the sparse opportunistic network in consideration [9]. This can be explained through the example of a State that qualifies as a reasonably sparse opportunistic network. Here the local public transport system (for instance buses) correspond to nodes that move in a restricted area of the city and when there needs to be an interstate movement designated interstate transport buses are used for the same purpose. Here the buses used may be different in terms of capacity and model suitable to catering the long-distance travel. This analogy can be derived from various examples including Vehicular delay-tolerant networks (VDTNs), DTNs for emergency communications among others [[10], [11], [12]]. The primary contributions of this papers are:

  • 1

    Leveraging social aspect in opportunistic-IoT communication through pervasive intelligence to enable sustainable computing. This enables a resource efficiency with respect to minimizing energy consumption, system overheads, buffer utilization and dropped messages, as

  • 2

    The proposed schemes makes use of a three-tiered clustering derived from the theory of Homophily and Small World theory, as explained in Section 3.1 and employs cluster-cloud generation using context-aware approach to facilitate routing.

  • 3

    This paper proposes a scheme to sectionalize the network topology based on node characteristics and employ grouping in intelligent transmission. The proposed scheme HiLSeR enables message routing using a combination of controlled-parameterized flooding and opportunistic sector-based decentralized transmission

  • 4

    The scheme makes use of the designated apex nodes at respective tiers to communicate with corresponding apex nodes which allows the scheme to skip several nodes in-between which further enhances energy efficiency and sustainable routing consideration.

The significance and contribution of this piece are manifold as this paper details not only for IoT devices but also introduces a novel opportunistic routing paradigm using many elements. The organization of this paper entails a total of VI sections. Under section II, we underscore the background and perform a literature review. Section III aims to explain the methodology of the paper followed by a comprehensive assessment across multiple performance parameters. The last section summarizes the paper by drawing relevant conclusions and subsequently, covers open challenges in this domain and the possibility of future work for improving the routing schemes.

Section snippets

Background and related works

In an opportunistic scenario, the primary aim is to institute reliable connectivity in an environment where there is a lack of end to end connectivity. As discussed above, the nodes have severe constraints with respect to resources and have to work in a scenario which has low density, restricted power and a range which is short. Given, this paper addresses OppIoT, there are special routing considerations that are to be addressed while making routing considerations. These include connectivity,

Motivation

As discussed above, a myriad of routing protocols has discussed this issue by using ubiquitous contextual and non-contextual information on the network to route messages. Every routing protocol is developed with the aim of minimizing the overhead, energy efficiency and ensuring security. Protocols cater to the above consideration by restricting flooding by taking the opportunity of contextual information regarding the network such as information pertaining to geography, social characteristics

Simulation setup

The proposed routing protocol is being implemented using the Opportunistic Environment abbreviated as ONE simulator. ONE simulator is a java based platform used for generating node movements by using various mobility models [32]. It enables visualization of both mobility and message passing in real-time in its graphical user interface (Table 3).

We are evaluating the proposed protocol parameters as per the following performance parameters [33]: Success Rate, Average Latency, Overhead Ratio,

Simulation results

The proposed protocol is compared with a group of six protocols which ranges from both extremes of the context-free and pervasive context-aware spectrum. The combination is chosen such that the comparison is holistic and includes different kinds of schemes ranging from nature-inspired routing, machine learning, K means clustering, Gaussian Mixture Models and Reinforced Learning for effective contrast. The protocols for this comparative assessment are Epidemic Routing - most elementary yet

Conclusion and future scope

In IoT, swift and effective data sharing has been a key area of consideration in the research community. The functionality of these devices has seen an increasing gradient while there has been a growing demand to reduce the space these devices occupy. Hence, there always have been efforts made to reduce the overhead, latency and accommodate a myriad of design constraints [34]. There has also been the concept of delay tolerance that is needed as the mobility of these devices has increased. This

Author statement

All authors provided critical feedback and equally contributed to shaping the research, analysis, and manuscript. Siddhant Banyal, Kartik Bharadwaj, Deepak Kumar Sharma, Ashish Khanna and Joel Rodrigues contributed to the design and implementation of the research, to the analysis of the results and to the writing of the manuscript.

CRediT authorship contribution statement

Siddhant Banyal: Conceptualization, Methodology, Formal analysis, Software. Kartik Krishna Bharadwaj: Conceptualization, Methodology, Formal analysis, Software. Deepak Kumar Sharma: Conceptualization, Methodology, Formal analysis, Software, Data curation, Validation, Supervision, Writing - review & editing, Project administration. Ashish Khanna: Data curation, Validation, Supervision, Writing - review & editing, Project administration. Joel J.P.C. Rodrigues: Data curation, Validation,

Declaration of Competing Interest

The authors report no declarations of interest.

Acknowledgements

This work is partially funded by FCT/MCTES through national funds and when applicable co-funded EU funds under the Project UIDB/50008/2020; and by Brazilian National Council for Scientific and Technological Development - CNPq, via Grant No. 309335/2017-5.

References (35)

  • T. Manzur

    Free space optical communications (fso)

  • H.F Sheikh, I. Ahmad, D. Fan, “An evolutionary technique for performance-energy-temperature optimized scheduling of...
  • L. Palen et al.

    Online forums supporting grassroots participation in emergency preparedness and response

    Commun. ACM

    (2007)
  • P. McDonald et al.

    Sensor network with delay tolerance (SeNDT)

  • A. Seth et al.

    Low-cost communication for rural internet kiosks using mechanical backhaul

    Proceedings of the 12th Annual International Conference on Mobile Computing and Networking

    (2006)
  • P. Juang et al.

    Energy-efficient computing for wildlife tracking: Design tradeoffs and early experiences with ZebraNet

    Proceedings of the 10th International Conference on Architectural Support for Programming Languages and Operating Systems

    (2002)
  • I. Ahmad, S. Ranka, “Handbook of Energy-Aware and Green Computing-Two Volume Set,” CRC...
  • Cited by (13)

    • A social relationship-based energy efficient routing scheme for Opportunistic Internet of Things

      2023, ICT Express
      Citation Excerpt :

      In [37] author proposed the data transmission based on social networks concept for opportunistic complex networks. Hierarchical machine learning based routing scheme for pervasive communication and resource efficiency is suggested in [38]. In [39,40] author proposed efficient routing technique based on trust value.

    View all citing articles on Scopus
    View full text