Relay node placement for building wireless sensor networks with reconfigurability provision

Abstract

Intelligent manufacturing needs the flexible and reconfigurable manufacturing systems to respond to dynamic market demands rapidly, which requires the underlying information systems (e.g., industrial wireless sensor networks) to provide reconfigurability. The Relay Node Placement (RNP) problem has aroused great attention due to its significant influence on network performance. However, existing works assume that the network configurations (e.g., transmit power, delay and reliability requirements) are fixed, and therefore cannot guarantee the reconfigurability of the deployed Wireless Sensor Networks (WSNs). To this end, this paper studies the RNP problem with reconfigurability provision. We propose an RNP framework with supporting network reconfiguration, and we also prove that this framework can ensure a deterministic approximation ratio and a tight time complexity. We also verify the efficiencies of the proposed algorithms through extensive simulations, and results show the superiorities of the proposed algorithms on deployment cost, end-to-end delay and reliability.

Introduction

In pace with the developments of wireless communications and microelectronics, Wireless Sensor Networks (WSNs) have gained tremendous advances, which can collect various information from the physical world. WSNs have been adopted in a wide range of applications, e.g., environment monitoring, battlefield surveillance, healthcare, smart grid and industrial automation [1], [2], [3].

Sensor Nodes (SNs) in WSNs are spatially distributed in target regions to gather certain information of interest, and the locations of SNs are generally predetermined and fixed [4], [5]. However, the Base Stations (BSs) are commonly installed near the boundaries of external networks with stable power supply [6]. Due to the limited communication range and power supply of SNs, Relay Nodes (RNs) are placed to build the connected network topology under various constraints (e.g., connectivity [7], [8], [9], [10], [11], [12], [13], [14], energy efficiency [15], [16], [17], [18], [19], [20], fault tolerance [21], [22], [23], deployment locations [24], [25] and delay [26], [27], [28], [29]). As network topology significantly impacts the performance of WSNs, the Relay Node Placement (RNP) problem, which is NP-hard [7], has aroused great attention [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24], [25], [26], [27], [28], [29].

Due to the merits of low-cost, handy-deployment and easy-maintenance, WSNs have been progressively used in factory automation [30], [31]. To build network topologies and transmit data reliably and in time, the Delay Constrained Relay Node Placement (DCRNP) problem has been studied in depth recently. Furthermore, the conception of intelligent manufacturing is proposed and has gained tremendous attention in these years. The implementation of this conception needs the manufacturing systems to provide flexibility and reconfigurability so as to respond to dynamic market demands promptly [32]. This requires the underlying information systems (e.g., industrial WSNs) to work correctly across different configurations and constraints [33], [34], [35].

Existing works regarding the RNP problem assume that the network configurations (e.g., transmit power, QoS requirements, etc.) are fixed and cannot be reconfigured during the entire network lifetime. As a result, these algorithms fail to support reconfigurability. To this end, this paper aims to build a network topology that is stably connected with respect to constraints across different network configurations through relay node placement, which is termed as the RNP problem with Reconfigurability provision (RRNP). In addition, the two-tiered architecture, in which SNs only send packets to their 1-hop neighbor RNs and the network connectivity relies on the mesh network composed of only RNs, brings many advantages such as energy efficiency and scalability. Therefore, this paper solves the RRNP problem in the two-tiered architecture [11].

We first propose a framework for the RRNP problem. This framework decomposes the RRNP problem into some subproblems that are instances of the DCRNP problems, and the RNs in each subproblem are placed for one network configuration. With the repetition of this framework, an adaptive WSN under different network configurations will be deployed. We also prove that the proposed framework promises an explicit approximation ratio and a tight time complexity through rigorous mathematical analysis.

Moreover, the objective of existing algorithms regarding the DCRNP problem is only to minimize the number of deployed RNs, providing no improvement on the network performances of end-to-end delay and reliability. Thus, a set of Differential Evolution (DE) based algorithms are presented to optimize the overall network performance (including end-to-end delay, reliability and the number of deployed RNs). Our main contributions are listed as follows:

• An RNP framework with supporting network reconfiguration is proposed. In addition, we prove that this framework ensures an explicit approximation ratio and a tight time complexity through rigorous mathematical analysis.

• To provide an efficient solution to each subproblem during the execution of the framework, the DE-based algorithms are presented. Different from existing algorithms, the DE-based algorithms not only try to save the deployment cost but also try to reduce the end-to-end delay and improve the Packet Reception Rate (PRR).

• We adopt a more realistic wireless channel model to estimate the wireless link quality during the node deployment instead of the ideal 0–1 disk channel model.

• Extensive simulations are conducted to evaluate the performances of the proposed algorithms. Simulation results verify the effectiveness of these algorithms.

The remaining parts of this paper are organized as follows: Section 2 reviews the related work regarding the RNP problem. Section 3 describes the system model and formulates the RRNP problem. Section 4 describes and analyzes the proposed framework for the RRNP problem. Section 5 presents the differential-evolution-based algorithm. Section 6 shows the simulation results. Finally, Section 7 concludes the paper.