SWARM-based data delivery in Social Internet of Things

Highlights

• To guarantee the connectivity among the SIoT objects and people, fault tolerance routing has been considered in this paper.

• We propose a bio-inspired particle multi-swarm optimization algorithm to construct, recover and select k-disjoint paths that tolerates the failure while satisfying quality of service (QoS) parameters.

• The validity of the proposed algorithm has been verified by comparing to the canonical particle swarm optimization (CPSO), and the fully particle multiswarm optimization (FPMSO).

Abstract

Social Internet of Things (SIoTs) refers to the rapidly growing network of connected objects and people that are able to collect and exchange data using embedded sensors. To guarantee the connectivity among these objects and people, fault tolerance routing has to be significantly considered. In this paper, we propose a bio-inspired particle multi-swarm optimization (PMSO) routing algorithm to construct, recover and select $k$-disjoint paths that tolerates the failure while satisfying quality of service (QoS) parameters. Multi-swarm strategy enables determining the optimal directions in selecting the multipath routing while exchanging messages from all positions in the network. The validity of the proposed algorithm is assessed and results demonstrate high-quality solutions compared with the canonical particle swarm optimization (CPSO), and fully particle multi-swarm optimization (FPMSO).

Introduction

Wireless sensor networks (WSNs) form the basis of the future networks and several applications of extreme importance in our present and future life in various aspects. “Social Internet of Thing (SIoT)” as shown in Fig. 1, is an exceptionally complex network model where varieties of components are deployed as consumer electronic devices that interact with one another in a complex way [1]. However, these devices operate under strict energy constraints, thereby making the dedicated energy budget for fault tolerant routing very limited [2]. The fault tolerant routing problem has received significant attention in the literature [3]. We believe that the emerging needs for SIoT applications, such as in smart homes, smart cities, and healthcare, will further increase the importance of fault tolerance in various aspects, due to its requirement constant mode of operation. Therefore special effort has been made to develop fault tolerance in routing [3].

WSNs often operate in an autonomous mode without human supervision in the loop [4]. Moreover, sensor nodes are often deployed in uncontrolled and sometimes even hostile environments [5]. Therefore, accurately predicting an optimal way to treat fault tolerance within a particular WSN routing approach is difficult because both the technology and envisioned applications for WSNs and SIoTs are changing at a rapid pace [6]. Given the limitations in power supply, the energy consumption in WSN is dominated by radio communication. The energy consumption of radio communication mainly depends on the number of bits of data to be transmitted with sensor network. In order to avoid the loss of significant data from the sensor node, the available communication energy is significantly lower than the computation energy [7]. For example, the energy cost of transmitting one bit is typically around 500–1000 times greater than that of a single $32$-bit computation [8]. Therefore, to recover from path failure, fault tolerance routing algorithms that require only a limited amount of communication energy regardless of any additional computational energy must be developed. Otherwise, any unpredictable events may cause the devices to fail, partition the network and disrupt the network functions.

These problems necessitate the development of fault tolerant routing approaches that require minimal additional computation regardless of any additional communication requirements hence as to construct or recover the selected path [9]. Basically, multipath routing protocols provide tolerance to failures and increase the network reliability [9]. The fault tolerant routing problem is often formulated as multi-objective optimization problem (MOP) to establish $k$-disjoint paths that guarantee connectivity even after the failure of up to $k$-1 paths. To have a more realistic analysis of our model, we formulate the strong fault tolerant routing problem as a MOP that is are treated simultaneously while being subjected to a set of constraints. Multiple objective may or may not be conflicting; therefore, multiple objectives cannot achieve their respective optimal values at the same time contrary to the problems of single-objective optimization (SOP) as illustrated in Fig. 2. Furthermore, a single globally optimal solution that is considered the best with respect to all objective functions.may be non-existent. Therefore, we should scalarizing an MOP means formulating a SOP such that optimal solutions are consider as Pareto-optimal solutions to the MOP. And the decision maker has to choose the best solution depending on the priorities of the QoS objectives to achieved.

Strong fault tolerance requires enormous computational efforts, which induce large control message overhead and lack of scalability as the problem size increases [10]. Solving these problems on an individual sensor node may require extreme memory and computational resources and yet produce average results [11]. We develop a bio-inspired particle swarm optimization (PSO) routing algorithm to achieve fast recovery from path failure. Owing to its simple concept and high efficiency, PSO has been actively utilized in these problems with promising results [12]. Despite its competitive learning performance, precisely solving the fault tolerant routing problem remains a challenging task for PSO due to premature convergence. Unfortunately, most premature convergence traps are caused by the rapid convergence characteristic and the diversity loss of the particle swarm, thereby leading to undesirable quality solutions [13]. We face another challenge in the ability of PSO to balance exploration and exploitation searches. Neither exploration nor exploitation must be overemphasized as exploration inhibits swarm convergence, while the exploitation tends to cause the particle swarm to hastily congregate without the feasible region that leads to premature convergence [14]. Motivated by these challenges, especially the actual connectivity model inside the WSN that has been integrated into SIoT, we propose two new routing algorithms, namely, fully particle multi-swarm (FPMS) and canonical particle multi-swarm (CPMS) which are based on multi-swarm computationally efficient alternatives to analytical methods that tolerate the failure of multipaths with reconstructed multipath and satisfy the quality of service (QoS) parameters in terms of energy consumption, delay, and throughput. Our contributions can be listed as follows:

• We develop a bio-inspired PSO routing algorithm to achieve fast recovery from path failure by attempting to extend an existing approach for finding an optimal solution in SOP for MOP. We define the objective functions and then optimize the effective values of these objective functions, which are computed at each sensor node that is selected to construct a $k$-disjoint multipath.

• We investigate the performance of the proposed multipath routing algorithms by comparing them with canonical particle swarm optimization (CPSO) [[15], [16]] to provide an alternative learning strategy for particles.

These algorithms are very similar to one another and only differ in a few details of their learning strategies in different situations with respect to convergence, exploitation, exploration, and jumping out of the basins of attraction of optimal solutions. Additionally, increasing the number of paths requires the exchange of more messages and additional communication overhead [17]. Therefore, by looking at the similarity and differences among these algorithms, we employ complex network connectivity to represent the population of swarm topology. We also adopt the multipath routing algorithm to balance the trade-off between fault tolerance and communication overhead by taking advantage of the combination of proactive and reactive routing mechanisms to exchange demanding information of calculation and maintain on every particle and records the objective function value for the selected paths. Afterward, the particles are adaptively increased or decreased and connected with their matching velocity to make a proper selection by considering the optimized objective function.

The rest of the paper is organized as followed. Section 2 introduces the related works. Section 3 introduces the concepts of the system model. Section 4 presents the problem solving mechanism. Section 5 discusses the simulation results. Section 6 concludes the paper.