Research of a resource-efficient, real-time and fault-tolerant wireless sensor network system

Abstract

Memory optimization, energy conservation, real-time scheduling and fault tolerance are the essential research topics in the wireless sensor network (WSN). In this paper, a memory-efficient, energy-efficient, real-time and fault-tolerant WSN system LiveOS is designed and implemented. Compared to the other systems, LiveOS has several typical features. Firstly, the optimized Rate-Monotonic Scheduling (RMS) scheduling algorithm is implemented. By doing this, the memory cost of the multithreaded real-time operating system (RTOS) can be 47.3% optimized if compared to the traditional single-core node. Consequently, the RTOS becomes appropriate to run even on the memory-constrained WSN nodes. Secondly, the new research approach that addresses the WSN challenges by combining both the multi-core hardware technique and the software technique is applied in LiveOS. By means of the multi-core hardware support, the energy cost of LiveOS can be 35% optimized compared to the traditional single-core platform. Moreover, the real-time scheduling performance and the fault-tolerant ability of the WSN nodes can be optimized significantly. Due to the above features, LiveOS becomes the WSN system which has small memory footprint, long lifetime, high real-time scheduling performance and upstanding fault-tolerant ability. The evaluation on the different prototype nodes proves that LiveOS is appropriate to run on the resource-constrained WSN platforms and is competent to execute the outdoor real-time WSN applications.

Introduction

Wireless Sensor Network (WSN) consists of a set of nodes which can monitor the physic environments and transmit the collected data cooperatively to the sink node (Akyildiz, Vuran, 2010, El Emary, Ramakrishnan, 2013). As the proliferation of the WSN technique, the WSN has played an important role in diverse application domains, including the precision agriculture, the smart care, etc. (Borges et al, 2014, Iyengar, Brooks, 2012).

Currently, several research challenges exist in the WSN. Firstly, the memory resources of the WSN nodes are constrained whereas the memory cost of the real-time multithreaded operating system (OS) is high. Therefore, the memory optimization to the real-time OS (RTOS) in the WSN becomes essential. With this optimization, the RTOS can become feasible to run even on the memory-constrained WSN nodes. Secondly, the WSN nodes are commonly powered by the energy-limited batteries, and the nodes are difficult to be recharged after being deployed outdoors. Therefore, the energy optimization becomes essential to the WSN nodes (Akyildiz and Vuran, 2010). With this optimization, the lifetime of the WSN nodes can be prolonged. Thirdly, the real-time reaction is required by many WSN applications, such as the industrial engine control and the contamination control. In these applications, the delay of the response to the time-critical tasks can cause disastrous results. Since the computation ability of the WSN microcontroller is limited, e.g., the working frequency of the AVR Atmega1281 microcontroller is only 16 MHz, the time-critical WSN tasks may not be completed within the deadline in case that the real-time scheduling overhead is high. Therefore, the optimization to the WSN real-time performance with which the high-overhead real-time tasks can still be schedulable becomes significant in the WSN. Finally, the WSN nodes are prone to be deployed in the harsh environments where the nodes are difficult to be recollected after being deployed. Therefore, the improvement of the fault-tolerant ability becomes critical to the WSN. With the tolerance to the faults, the maintenance cost during the post-deployment process can be decreased.

In past years, many research works have been done to address the challenges above. To optimize the memory cost of the multithreaded system, the stack-analysis approach (AbsInt Corporation, 2007, Torgerson, 2005) and the cooperative scheduling mechanism (Keen et al., 2003) were proposed. To conserve the energy cost, the data prediction (Jain et al, 2004, Kanagal, Deshpande, 2008), the code compression (Caione et al, 2012, Srisooksai et al, 2012), the data aggregation (Maraiya et al, 2011, Xiang et al, 2011) and the topology control (Aziz et al, 2013, Li et al, 2013, Üster, Lin, 2011) mechanisms were applied. To improve the real-time scheduling performance of the WSN system, the real-time preemptive algorithm (Buttazzo, 2005), the resource pre-reservation mechanism (Eswaran et al., 2005), the time-efficient network protocols (Aissani et al, 2013, Mouradian et al, 2014, Xia et al, 2013), the data aggregation (Maraiya et al., 2011) and compression mechanisms (Szalapski and Madria, 2013) were researched. To improve the fault-tolerant ability of the WSN system, the multipath routing (Chanak and Banerjee, 2013), the data aggregation (Maraiya et al., 2011) and the topology management (Younis et al., 2014) mechanisms were implemented. With the above mechanisms, the memory cost, the lifetime, the real-time performance and the fault-tolerant ability of the WSN system can be optimized significantly. However, these optimizations are still not sufficient for the proliferation of the WSN. Currently, the investigation of the new design concepts and the new research approaches which can address the above challenges further is still essential.

In this article, a new WSN system LiveOS is designed and implemented. Compared to the other WSN systems, LiveOS has several typical features. On the one hand, the new optimized rate-monotonic scheduling (RMS) mechanism (Liu and Layland, 1973) is realized. With this mechanism, the memory cost of the multithreaded RTOS can be decreased significantly by comparing to the traditional multithreaded WSN OS. As a result, the RTOS becomes feasible to run even on the high memory-constrained WSN platform. On the other hand, the new research approach, which addresses the WSN challenges by combining both the software technique and the multi-core hardware technique, is achieved in LiveOS. By means of the multi-core architecture, the energy conservation, the real-time performance and the fault-tolerant ability of the WSN nodes can be optimized effectively. With the implementation of the above mechanisms, the WSN memory and energy constraint challenges can be addressed, and the high-overhead real-time tasks are prone to be schedulable. Moreover, the availability of the WSN nodes can be improved.

The structure of this article is organized as follows: In Section 2, the related works about the WSN memory optimization, energy conservation, real-time scheduling and fault tolerance are presented. In Section 3, the LiveOS optimized RMS scheduling mechanism is discussed. With this mechanism, the memory cost of LiveOS can be reduced significantly. In Section 4, the LiveOS multi-core energy conservation scheme is introduced. With this mechanism, the lifetime of the LiveOS node can be prolonged by comparing with traditional single-core WSN node. In Section 5, the LiveOS multi-core real-time scheduling mechanism is presented. With the implementation of this mechanism, the real-time tasks can be schedulable even if their overheads are high. In Section 6, the LiveOS multi-core fault-tolerant mechanism is investigated. By doing this, the availability of the WSN nodes can be improved. Finally in Section 7, the conclusion and the ongoing works are proposed.