Self-organizing Approach for Real-time Data Fusion in Dense

0 downloads 0 Views 492KB Size Report
Departamento de Ciências da Computação (PGCC) – UFSC, Brasil. 3. Pós-graduação em Engenharia Elétrica (PGEEL) – UFSC, Brasil. 4. Pós-graduação em ...
11th Brazilian Workshop on Real-Time and Embedded Systems

Self-organizing Approach for Real-time Data Fusion in Dense Wireless Sensor Networks A. R. Pinto1,3, Marcos Camada4, Paulo Portugal1 M. A. R. Dantas2, Carlos Montez3,4 1

ISR/IDMEC/FEUP – Universidade do Porto, Portugal Departamento de Ciências da Computação (PGCC) – UFSC, Brasil 3 Pós-graduação em Engenharia Elétrica (PGEEL) – UFSC, Brasil 4 Pós-graduação em Engenharia de Automação e Sistemas (PGEAS) – UFSC, Brasil 2

{arpinto,pportugal}@fe.up.pt, [email protected], {mcamada,montez}@das.ufsc.br

Abstract. Wireless Sensor Networks (WSN) are usually composed for a huge number of tiny nodes deployed in an environment that has to be monitored. Due to high probability of faults in communication level and in the nodes itself, network topology becomes very dynamic and unpredictable. Several challenges are faced in these networks like real-time and quality of service. Moreover, it is very difficult to fix the network due to environmental issues or the huge number of nodes. This way self-organizing and self-organizing approaches are required to maintain the network services. In this paper a selforganizing approach for real-time data fusion applications is presented. Our approach showed that is possible to achieve the required levels of communication efficiency or quality of service required by system administrators.

1. Introduction Wireless Sensor Networks (WSN) are usually composed for a huge number of tiny nodes deployed in an environment that has to be monitored. Nodes have processor, memory, sensors, battery and wireless communication module. The high interaction degree with the environment of these networks implies requirements that were not aimed in traditional distributed applications. For instance, energy efficiency, real-time, fault tolerance and non-controlled environment issues [11]. Due to high probability of faults in communication level and in the nodes itself, network topology becomes very dynamic and unpredictable. In this way, the strategy of deployment of a huge number of low-cost sensors – compared with the use of few expensive and reliable sensor nodes – offers several advantages: high robustness against failures, uniform covering e easy of deployment [1]. There are several data fusion approaches generally target increasing sensor reading dependability, enhancing energy efficiency and increasing network lifetime [2,7,8,9,10]. Data sensed are sent to base station, where they are fused in order to obtain some useful information. This way, that is not necessary trust in just one sensor anymore, but in a set of collected samples. Thus, even if some nodes are faulty, decisions can be based on various sensors of the network. This is one of the most important characteristic of these

45

46

11th Brazilian Workshop on Real-Time and Embedded Systems

networks: network and data dependability are not based on just one sensor but on a set of them. WSN are dense wireless networks. They have real-time requirements and they can be used to monitor hazardous and/or inaccessible areas. Thus, self-management characteristics are necessary, because it is extremely difficult to manage a huge number of nodes that use wireless media and that have many conflicting goals (energy efficiency, self-organization, real-time requirements and fault tolerance). Selfmanagement is one of the most important characteristics of autonomic computing – that was proposed in 2001 – these approaches aim to automate the system administration tasks. Computing system components are able to self-organize e self-optimize according to global goals that are dictated by system administrator or system user [6, 11]. In this paper, a monitoring application with homogeneous signal is proposed, where there are N slave nodes (that sense and send collected data to a master node). Master node is responsible for periodically fuses data, based on data received from slave nodes. Thus, one using mobile devices is able to monitoring inaccessible or hazardous areas, and to take decisions based on this information. Master node does not have information of how many slave nodes are active. This way, an autonomic approach based on a relation between Quality of Fusion and Communication Efficiency is proposed. This approach tries to achieve a certain level of communication efficiency, even facing communication faults and random topologies. The main goal of this approach is to reach a communication efficiency dictated by system manager. Simulation results shown that the proposed approach could reach the metrics imposed. Related theory of WSN, data fusion and IEEE 802.15.4 are introduced in section 2. The proposed scenario is described is section 3. The autonomic approach and simulation results are presented in section 4. In section 5 a comparison with related works is described. Final remarks are presented in section 6.

2. Background 2.1. IEEE 802.15.4 Standard The main target of the ZigBee technology is to enable the setup of dense low-power wireless networks. There are many products like Mica Motes, which use this network technology in order to achieve longer lifetimes in WSN applications [13]. The architecture of ZigBee is based on the Open System Interconnection (OSI) standard, where a subset of layers is implemented. ZigBee adopts IEEE 802.15.4 standard that defines two layers: physical and medium access layers [3]. The physical layer can operate in two frequencies 868/915 MHz or 2.4 GHz (16 channels and 250 Kbps of maximum transmission rate). IEEE 802.15.4 medium access sublayer uses the CSMA-CA medium access mechanism. This is a low-rate, low-cost and low power technology that allows the implementation of self-organizing and flexible topologies [3]. IEEE 802.15.4 was proposed in 2003 and is becoming a de facto standard for low power and low rate wireless networks. IEEE 802.15.4 MAC supports two types of operational modes that can be selected by a central node called PAN coordinator: (1) beaconless mode, a non-slotted CSMA/CA;(2) beacon mode, where beacons are sent

11th Brazilian Workshop on Real-Time and Embedded Systems

periodically by PAN coordinator. In this case, nodes are synchronized by a superframe structure. IEEE 802.15.4 operating in beacon mode presents a “Contention Free Period”, where a “guaranteed slot time” is used. ZigBee Alliance is currently working in application and network layers (clock synchronization is also considered as an optional superframe structure). A ZigBee Network can enable 65000 nodes, based on its address scheme. Three types of topologies are supported: star, mesh and cluster tree. The Star Topology is considered the simplest scheme, where nodes achieve to communicate with each other in just one hop. Therefore, mesh and tree topologies need to use routing techniques in order to organize nodes. Moreover, a tree topology is composed for several clusters of star topologies, interconnected by its cluster-heads. However, routing and organizing costs in tree or mesh networks are so expensive that the use of a star topology is usually more suitable [13]. 2.2. Wireless Sensor Networks Wireless sensor Network is a general term that covers many variations in composition and deployment [4]. These networks are composed for large number of small nodes consisting of sensor unit, wireless antenna adapter, processor, memory and battery. The nodes are able to sense, process, and communicates data with each other, handling them typically over a radio frequency channel in direction to an interested end user. Sensor nodes can be homogeneous or they may have some special nodes with special characteristics. A base station is necessary to collect and process data sensed by ordinary sensor nodes. Resources are restricted in a WSN node. Nodes usually presents limited processing capacity, small memory and limited energy resource. Furthermore, most networks have high density of nodes, wireless communication and hardware components are prone to failure and topology is very dynamically [1,4,11]. These constraints impose a lot of challenges in a sensor network application. The sensors in these networks usually operate in a collaborative fashion, combining data sensed by multiple sensors [2,7,8,9,10]. It is necessary for reliable execution of such tasks due to a variety of reasons, including limited information gathered by each node, variability in operating conditions, and node failure. This way, the main goal of a data fusion approach is combine data sensed by multiple sensors in order to improve quality of decisions in base station. 2.3 Data Fusion Wireless sensor networks are often deployed in an environment in order to detect some physical phenomenon signal. The detection can be usually done in a collaborative fashion, combining data sensed by multiple sensors [10]. Exchange of information between nodes is necessary for reliable execution of such tasks due to a variety of reasons, including limited information gathered by each node, variability in operating conditions, and node failure [8]. This way, the main goal of a data fusion approach is combine data sensed by multiple sensors in order to improve quality of decisions in a sink (i.e. base station).

47

48

11th Brazilian Workshop on Real-Time and Embedded Systems

Collaborative detection can be done in a serial or parallel fashion, considering node communication. Parallel data fusion is executed when all nodes send their sense data to a base station. On the other hand, serial data fusion uses routing techniques in order to collect data through the network [6]. Hybrid data fusion approach organizes the data fusion process in clusters or tree-scheme. Therefore, parallel data fusion occurs inside a cluster. On the other hand, cluster heads (nodes that perform data fusion) exchange data between them in a serial fashion. Data fusion must be considered when raw data is sent to a node that executes fusion process. This approach is usually used when data is correlated. On the other hand, if readings are independent, just probabilistic information are exchanged. According to [5], this technique should be called “decision fusion”, but in this paper we will use the expression “data fusion” in both situations.

3. Adopted scenario This section presents the scenario adopted in this present work. The same scenario – based on a master-slave discipline over a star topology – was previously detailed in [13]. The used communication model considers one master node (base station) and N slave nodes (Figure 1), where the slave nodes periodically sense scalar data. The signal is considered to be homogeneous in the monitoring area. Data collected by slaves is sent to the master node that performs the data fusion. All the slave nodes reach the master using just one hop. That is, a parallel data fusion is performed in master node.

Figure 1: System architecture.

The concept of monitoring session is adopted. A monitoring session is a time interval where all slave nodes send periodically sensed data to the master node. A session S is composed for N TS rounds with the length R. Therefore, it is composed for 0,R,2R,3R, ..., (N-1)R rounds. The round concept is used to synchronize nodes, and it also represents the periodicity of the data fusion task. On each round, a slave node can send zero or one message M containing the sensed data to the master node. All slave nodes are synchronized by the WSN round concept. Each message M sent by a slave node has an absolute deadline D that is the maximum time interval within

11th Brazilian Workshop on Real-Time and Embedded Systems

which it must be delivered to the master node. Otherwise, it will no longer be useful for the data fusion task. This absolute deadline is computed based on a relative deadline d. We considered a homogeneous architecture where all slave nodes have the same relative deadline. This relative deadline value is sent by the master node in the beginning of the session. The absolute deadline of a slave node at round n is D=nR+d, where R is the round length. The master node performs a data fusion operation considering just the messages that arrived on time. In this work, the master node just fuses data that arrived within the same round. That is, the relative deadline of a message sent in round n is always 0

Suggest Documents