Investigation of Architecture and Essential Technologies of Wireless Sensor Network

The Wireless Sensor Network (WSN) is a distributed Ad-hoc network that consists of numerous and ubiquitous mini sensor nodes with the capabilities of wireless communicating and computing. The WSN is an “intelligent” system that enables self-fulfillment of assigned tasks according to specific environments. It integrates the technologies for wireless communication and distributed information processing as well as for sensors and Micro Electro-Mechanical System (MEMS). The WSN is different from either traditional wireless networks (such as 2G and 3G mobile communication networks) or the wireless Local Area Network (LAN). The design goal of the traditional wireless networks is to optimize the utilization of bandwidth in the high-speed mobility conditions by the strategy of resource management while offer certain Quality of Service Service (QoS). The wireless LAN enables a heavy data throughput through routing optimization and high-efficient resource management protocols. Moreover, it supports communication of static and low-speed nodes. However, the WSN aims at transporting data. It is not required to support a high-speed data transmission rate. In addition, most nodes of the WSN are static. These nodes usually work in atrocious or even dangerous environments where humans are unable to arrive. Therefore, the power of these nodes is irreplaceable. The core of the design of a WSN is an efficient strategy for prolonging the life span of the network.

    No standardization organizations have defined the WSN architecture yet. Since the composition, protocols and architecture of the WSN are tied up with upper applications, the network architecture for one application is somewhat, or even totally, different from that for another application.

1 Architecture of WSN
A WSN includes distributed sensor nodes, transceivers and interfaces between the Internet and users. The sensor node is composed of the sensor unit, processing unit, communication unit and power unit. All the sensor nodes consist of the basic and core unit of the network[1]. Generally, adopting the layered structure, the WSN is divided into the physical layer, data link layer, network layer, transmission layer and application layer. The physical layer may use advanced Radio Frequency (RF) transmission technologies such as Orthogonal Frequency Division Multiplex (OFDM), Ultra Wide Band (UWB), Multi-Input Multi-Output (MIMO) and even Code Division Multiple Access (CDMA). It can also adopt traditional infrared transmission technologies. The main task of the data link layer is to guarantee the exactness of data transported by the physical layer, while improving the frequency utilization of the system. The functions of the network layer include packet routing, congestion control and more. The transmission layer offers reliable transmission service with reasonable overhead. The application layer provides various specific value-added services, as well as the functions of time synchronization and node positioning.

    The control management technologies are used for the layered protocol structure of the WSN. However, without standardized protocols, the protocol structures are omnifarious. A proposed cross-layer management mechanism of the WSN is shown in Figure 1.

    In order to save energy and prolong the life span of the battery, the nodes in a WSN will automatically put themselves into the sleep mode when no wireless data are being sent, received or transferred. This mode change may change the routing topology of the system as well. It is necessary to rebuild the route. Therefore, before the node goes into sleep mode, it is recommended to check whether the routing topology will be changed. If the change of the routing topology will happen, the mode change of the node shall be prohibited. It is necessary to perform this check to avoid the instability of the network topology.

    Except that the nodes of WSN tend to make the conversion between the active and sleep modes, its radio links suffer from the common influence of time-selective fading, frequency-selective fading, space-selective fading, interference and noise. Therefore, it is necessary to report the quality of signals at the physical layer to the Medium Access Control (MAC) layer in time. Besides, the MAC layer needs to exchange information with the network, transmission and application layers as well.

    Traditional network management based on layered protocols cannot meet the demand of WSN. Therefore, the advanced cross-layer network management technology becomes the guarantee of efficient operation of a WSN. The control technologies that match the cross-layer protocol structure include power management, QoS management, radio resource management and mobility management. As a distributed network, the WSN should have both security management and network management at each layer to ensure the security of protocols at every layer.

    It is necessary for the nodes to offer various radio services with specific QoS requirements to ensure the reliability of data transmission. The QoS technology should have full flexibility and self-adaptability.

    Energy management plays a very important role in the WSN. The WSN nodes request low power consumption. Power is one of the most important resources for the WSN, which is different from traditional wireless communication networks. Therefore, power control is involved in distributed wireless networks. Benefits of the power control include:

    (1) Adopting small transmitting power as possible or reducing dispensable power transmitting can not only prolong the life span of batteries, but also augment the network capacity.

    (2) Reducing the transmitting power can decrease link collisions.

    (3) When the traffic load of the system is high, low transmitting power helps obtain smaller end-to-end time delay. However, when the traffic load is low, it is high transmitting power that helps get smaller time delay. The power control system determines power allocation according to the quantity of traffic loads.

    (4) The system usually integrates the routing technology for the network layer to make optimization design.
Wireless resource management includes power control mentioned above, as well as capacity and load management and the mechanism for distributing and scheduling the resources. On one hand, wireless resource management helps meet various QoS requirements. On the other hand, it ensures the resource utilization with high efficiency. Advanced technologies for the physical layer such as OFDM, MIMO and UWB are adopted for the bottom layer of the WSN. Therefore, the algorithms of wireless resource management are required to change from traditional one-dimensional resource management to the time-space or time-frequency two-dimensional, or to the time-space-frequency
three-dimensional. Besides the operation of basic resource management such as transmitting power control and allocation, rate allocation and rate adjustment, the management should closely combine with the technologies adopted for the physical layer. For example, if the MIMO technology is used, antenna selection and space division multiplexing need to be managed. However, sub-carrier allocation is necessary when the OFDM technology is adopted.

    With mobility management, some WSN nodes may move at a low rate, which broadens the application scope of the WSN. In order to support the mobility, the followings are required. The physical layer is required to make measurement; the MAC layer to do handover and control; the network layer to adjust and maintain routes; and the upper layer to implement data cache and congestion control.

2 MAC Protocols
The data link layer is the core of the WSN to guarantee correct data transmission. The common MAC protocols for it include Sensor-Medium Access Control (S-MAC)[2] and Timeout-Medium Access Control (T-MAC)[3]. They use the Carrier Sense Multiple Access with Collision Detection (CSMA/CD) protocol to avoid collisions when sending data. S-MAC divides time into frames. The length of the frames is flexibly defined by application programs.
Each frame is further divided into the duty cycle and the sleep cycle. In order to support data burst best, the length of the duty cycle is adjustable. When sending data, S-MAC adopts a RTS/CTS/DATA/ACK mechanism to avoid collision. Good scalability and strong adaptability to changes of network topology are the strengths of this protocol. However, it has weaknesses in complicated protocol implementation and large overhead. In addition, since usable resources are relatively few for the WSN, low frequency utilization of this protocol becomes a bottleneck of its application in the WSN. Similar to S-MAC, T-MAC divides time into frames. Each frame is also divided into the duty cycle and the sleep cycle, while the length of the duty cycle is adjustable. Differently, the length of the frames is fixed. In addition, T-MAC defines five events and the Adaptive Timeout (TA). The five events are the following: the frame timeouts; a node receives data; data transmission collisions; a node completes data sending; adjacent nodes complete data exchange. If the TA detects no happening of these five events at a channel, the channel is considered idling. At this time, the node closes the RF module to save energy, and the channel goes into the sleep cycle. In order to reduce the power consumption for detection of an idle status, S-MAC and T-MAC has been improved and developed into WiseMAC[4] and Berkeley-MAC[5]. The later two protocols adopt detection technology with low power consumption to detect the idle status. Besides, they use CSMA when sending data. All the MAC protocols mentioned above are complicated and difficult to be implemented. In addition, they cannot match the characteristics of wireless distributed networks. However, the following three MAC protocols developed by IEEE are well applicable to the WSN.

2.1 IEEE 802.15.4 MAC Protocol
As a mature protocol for Wireless Personal Area Network (WPAN), IEEE 802.15.4 can be implemented with little complication, extremely low cost and small power consumption. Moreover, it supports low-rate data transmission between inexpensive fixed, portable and mobile devices[6]. IEEE 802.15.4 provides two physical layer options (at 868 MHz/915 MHz or at 2.4 GHz). The coordination of the physical and MAC layers broadens the network application scope. The two physical layer options adopt the DSSS technology to reduce the cost of digital integrated circuits. Besides, they use the same packet structure to realize operation with a short period and low power consumption. The physical layer at 2.4 GHz has a data transmission rate of 250 kb/s, while that of 868 MHz/915 MHz has 20 kb/s and 40 kb/s respectively.

    IEEE 802.15.4 adopts carrier detection and collision avoidance. Nodes periodically detect channels, receiving the beacon frames from the channels. A node will sleep once there is no data sending and receiving (as shown in Figure 2). The network coordinator will cache the data sent to the node in sleep, and then periodically send beacon frames to the node. These beacon frames carry the destination information of the data. When the node in sleep finds there is information sent to it, it sends a poll frame to the network coordinator to indicate that it can receive the data. After receiving the poll frame, the network coordinator sends an ACK frame to the originating node, and then sends the cached data. The destination node will send an ACK frame to the network coordinator once it receives the data.

2.2 IEEE 802.15.3 MAC Protocol
IEEE 802.15.3 is a new protocol for WPAN[7]. Its physical layer adopts Multiband Orthogonal Frequency Division Multiplexing (MB-OFDM) UWB and Direct Sequence-Code Division Multiplex Access (DS-CDMA) UWB technologies. It allows 245 wireless user devices in a range of several centimeters to 100 meters to access a network at a top rate of 55 Mb/s simultaneously. It provides fixed and mobile equipment with high-rate wireless access at
2.4 GHz. IEEE 802.15.3 specifies five original data rates: 11 Mb/s, 22 Mb/s, 33 Mb/s, 44 Mb/s and 55 Mb/s.  The rates will influence transmission distances. For example, the distance of transmission at 55 Mb/s is 50 m, while at 22 Mb/s is 100 m.  High rates (such as 55 Mb/s) can support low-delay multi-medium access and big-file transfer service, but low rates (such as 11 Mb/s and 22 Mb/s) offers long-distance connection between audio equipment.
IEEE 802.15.3 has all the elements required for QoS guarantee. Moreover, it uses Time Division Multiplex Access (TDMA) technology to allocate channels to avoid collisions.

    IEEE 802.15.3 just defines the protocols for the physical and MAC layers. Its MAC protocol comes from the IEEE 802.11 MAC protocol for Wireless Local Area Network (WLAN). Therefore, it is based on the Ad hoc structure and still has a trace of the star topology (as illustrated in Figure 3). PicoNet is the basic unit of a WSN based on IEEE 802.15.3. Its core equipment is called PicoNet Coordinator (PNC). The PNC is responsible for offering the synchronized clock, QoS control, power saving mode and access control. PicoNet, a kind of the Ad hoc network, exists only when there is a demand for communication. It will disappear with the end of the communication. Other equipment in PicoNet is the communication node (DEVs) of the WSN. Data exchange in a WSN is made between the DEVs, but control information of the network is sent by PNC.

    Figure 4 shows the super-frame structure of a WSN based on IEEE 802.15.3. One super frame consists of three parts:

    (1) Beacon: having information about clock allocation and communication management

    (2) Contention Access Period (CAP): used for exchanging commands and asynchronized data

    (3) Channel Time Allocation Period (CTAP): including several Channel Time Allocation (CTA) units, some of which are Management Channel Time Allocation (MCTA) units.

    A PicoNet emerges at the time when the PNC starts to send beacons. The beacons carry information about this PicoNet. Even if there are no communication nodes, a PNC that has been sending beacons can be regarded as a PicoNet. When a PicoNet is emerging, the PNC first finds a usable channel and sends a beacon frame to make sure the channel is free. Then the PicoNet is established at this channel. After a PicoNet is established, the PNC can still be changed by handover control. However, IEEE 802.15.3 doesn´t support the function of integrating two PicoNets into one.

    The PNC implements allocation of the air resources by sending beacons. The beacons carry network control parameters (such as network synchronization and the maximal transmission power), information about allocation of channel time slot, indication information aiming at every service flow transmitted by super frames, and more. CAP uses the MAC mechanism for CSMA/CA. In the CTAP, allocation may be implemented by basic TDMA. MCTA can also use TDMA to allocate time. Besides, equipment is permitted to share the MCTA (based on the ALOHA protocol).

    The IEEE 802.15.3 MAC protocol evolves from IEEE 802.11 MAC. Though data are directly transferred among wireless sensors (mainly referring to UWB equipment), center control is still necessary. This star topology is applicable to the WSN with a PC center (with strong processing capability and large storage room). However, it fails to well support Consumer Electronics (CE) and communication equipment that require simpler and mobility-supported connection. New protocols for the MAC layer keep emerging. On one hand, the IEEE 802.15 working group planned to initiate a study on new MAC protocol in IEEE 802.15.3b. On the other hand, MB-OFDM has great support and promotion from equipment vendors and research institutes, and Multiband OFDM Alliance (MBOA) is making its own MAC protocol. In order to support CE and communication equipment better, the MBOA MAC protocol will support the center control structure and the distributed network topology as well. It is expected to have the following features:

    (1) Any equipment is allowed to establish a network.

    (2) Power control is implemented to reduce interference.

    (3) The protocols for access and data transfer are simple.

    (4) A wireless link can be quickly set up or cut off (in less than one second).

    (5) Installation is simple (with zero setting).

    (6) Convergence and split of networkis supported.

    (7) Cross-network mobility is supported.

    (8) Mutual coordination between networks is supported.

    (9) More electricity is saved.

    (10) Both synchronized and asynchronized services are supported.

    (11) Wireless mesh topology is supported.

    The unit of the MBOA-based WSN is the Beacon Group (BG). All super frames of the WSN nodes have a uniform length but the frame structures of the BGs differ. The WSN node can roam from one BG to another by changing its frame structure. Moreover, the node can simultaneously follow two frame structures and become a common member of the two BGs. In addition, the node can transfer data between two relay stations. The wireless mesh network is accordingly established.

    In the super-frame structure of the MBOA MAC protocol, every device (except equipment in sleep mode) will send beacons. The beacon period is changeable to accept different quantities of equipment. A device will look for the beacons of other devices first. If it finds no beacons available, it will create a new beacon. However, if the device finds one available, it will join to it and keep using the same time slot. Equipment implements different QoS by
labeling resource reservation with different levels in beacons. Asynchronized data adopt the contention access mechanism with multiple priorities.

2.3 IEEE 802.16 Mesh MAC Protocol
The wireless mesh network was originally established to meet demands on battlefield communication of US Military. With a form of mobile Ad hoc network, it was dedicated to meet the demands for broadband data transmission, end-to-end IP support, voice and video support. In addition, it shall be positioning independent of Global Positioning System (GPS) but with the precision of GPS in military communications. The mesh network features robustness, flexibility and high-rate data transmission.

    The wireless mesh system was proposed for expanding the coverage of centralized control networks. The multi-hop technology similar to mobile Ad hoc is adopted to organically combine centralized and distributed control
technologies. As shown in Figure 5, scheduling in the wireless mesh network may be centralized, distributed, or a centralized-distributed mixture. In Figure 5, both the wireless Ad hoc network and WSN equate to the undistributed wireless mesh network.

    The structure of wireless mesh frames adopts time division multiplexing, as shown in Figure 6. A frame consists of a control sub-frame and a data sub-frame. The control sub-frame is divided into the network control sub-frame and the scheduling control sub-frame. The network control sub-frame implements network control by sending the access message and the deployment message. Its function is to create and maintain the consistency of different systems, including managing the connection of network nodes, selection of links and control information of nodes and links.
The scheduling control sub-frame does carry information of both the centralized and distributed scheduling. It is used to decide the allocation of resources on a link, and implement coordinated scheduling of data transmission between systems. The data sub-frame is used to transmit users´ data. Moreover, distributed non-coordinated scheduling messages may be sent under the condition that no collisions with coordinated scheduling take place[8].

    IEEE 802.16 recommends that the physical layer of the mesh network adopts OFDM technology. Each frame is labeled with a 12-bit frame number for addressing. It is divided into 256 mini-slots. The network control
sub-frame is periodically produced, while all the other frames have a scheduling control sub-frame. The length of the control frame is fixed, which is arranged by the mesh base stations, and indicated through the variable L. Each control sub-frame has L×7 OFDM signs. In order to send control and scheduling messages, the protocol divides the control sub-frame according to the sending chance. In a network control sub-frame, the first group of seven OFDM signs is the sending chance for Mesh Network Entry (MSH-NENT), used for network access. The following (L-1) ×7 signs are the sending chance for Mesh Network Configuration (MSH-NCFG), used for notification of network configuration messages. In a scheduling control sub-frame, the number of distributed scheduling messages is decided by the mesh base station, and presented by the variable N.

    According to the network description, the first (L-N ) ×7 signs of the sub-frame are used for sending Protocol Data Unit (PDU) bursts containing either the Centralized Scheduling Configuration of Mesh Network (MSH-CSCF) or Centralized Scheduling Channel of Mesh Network (MSH-CSCH). The other part of the sub-frame is used for sending the bursts containing Distributed Scheduling Channel of the Mesh Network (MSH-DSCH). If no scheduling collisions are indicated in the control sub-frame, distributed scheduling messages that use long preamble probably appear in the data sub-frame. The Quadrature Phase Shift Keying (QPSK) 1/2 modulation and encoding are used for all the messages sent by the control sub-frames.

    In IEEE 802.16, MSH-NENT messages are used to provide network access for new nodes. Before a new node becomes a fully functional member of the network, the network access scheduling protocol requests the upper-layer protocol to offer an unreliable mechanism for the access of the network entry slot. This will make the new node communicate with fully functional members in the network. A node in the network entry slot sends a MSH-NENT message in the following steps:

    (1) After the destination responsible node sends a MSH-NCFG message, the new node randomly sends a message in the idle network entry slot next to the sending chance of MSH-NENT. Contention and collision exists here. The initiative message contains the MSH-NENT packet of the request Information Unit (IE). If the MAC address of the responsible node in this message is the 16-mechanism value 000000000000, this responsible node is usable.

    (2) After the responsible node broadcasts the MAC address of the new node by the MSH-NCFG message, the new node may sends its MSH-NENT message in the sending chance next to MSH-NENT.

3 Routing Protocols for WSN
Presently, there are several routing protocols for Mobile Ad-hoc Networks (MANET). However, with regard to the characteristics of the WSN, these protocols cannot be used for the WSN directly. The task of a routing protocol for the WSN is to establish a route between a sensor node and a central transfer node to ensure reliable transmission of data. The network resources of a WSN are limited. Therefore, it is impossible for a node to make complicated computing. Small cache capacity doesn´t allow the node to store too much routing information. In addition, too much routing information exchange between nodes is not permitted.

    To be precise, the wireless routing protocols are divided into the table-driven and the need-driven. The table-driven routing protocol features continuous update. Each node in this kind of protocols maintains one or more tables to store routing information, and will broadcast update messages when the network topology changes. The table-driven routing protocol is composed of Destination Sequenced Distance Vector (DSDV), Clusterhead Gateway Switch Routing (CGSR) and Wireless Routing Protocol (WRP). Comparatively, the need-driven routing protocol is a dynamic protocol, adopting the need-driven routing algorithms such as the Ad Hoc On Demand Distance Vector (AODV) algorithm, Dynamic Source Routing (DSR) algorithm and Temporally Ordered Routing Algorithm (TORA). When a node needs a route to the new destination node, it has to wait until the route is found.

    In a WSN, nodes look for addresses according to the attributes of data. The data detected by the nodes are usually sent to the central transfer node first. Moreover, the nodes have weak mobility. According to these characteristics of the WSN, the need-driven routing protocol has the priority. In order to meet the requirements of energy management, a routing protocol must ensure the minimal energy consumption and loss based on the premise that QoS is guaranteed.

    The basic goal of the routing protocols for traditional wireless Ad hoc distributed networks is to ensure the normal operation of the networks independent of infrastructure. However, the routing protocols for the WSN aim at offering reliable multi-node data transmission. Therefore, the WSN working by fixed-node multi-hop routing need no complicated distributed routing algorithms. However, flexible routing is still necessary for timely routing change corresponding to the change of the link status or flow mode.

    Currently a number of routing protocols for Ad hoc networks work by looking for a route with the fewest hops. Their great strength is the simplicity. It is easy to calculate hops and find a route with the fewest hops between the source and destination nodes, once the network topology is built. In addition, hop calculation does not ask for additional parameters. However, such routing does not take the packet loss rate and bandwidth into consideration. This is its biggest weakness. The simple consideration of the number of hops is not enough for finding an effective
delay-throughput-reliability-balanced link. A route with the fewest hops is not always the route with the largest throughput, because it probably contains some wireless links far away and/or the packet loss rate is high. For example, a reliable route with two hops and a high data rate performs better than one with one hop, but with low packet loss and data rates.
In order to find effective routing algorithms for the WSN, a cross-layer design is the good way, which combines routing selection in the wireless mesh network with measurement of the physical layer and wireless resource management of the MAC layer. Such a design aims at wireless routing with the most energy saving, least interference and fewest hops as possible, finally improving system performance.
4 Capacity and Time Delay of
    WSN
In a WSN, the sensor nodes need to communicate with the central transfer node. They connect to the central transfer node by a single hop or multi-hops. Therefore, all the nodes in the covered range of the central transfer node must be controlled by the MAC protocol and contend for the channel capacity. The system capacity is effected by multiple factors including the number of nodes (n), covered area (A), transmission power (p) and the distribution of the central transfer node and sensor nodes. On one hand, the addition of nodes not only increases signal interference between nodes, but also causes collisions of the bottom links. Therefore, the throughput of a single node to the number of nodes n is approximately 0(1/nα). When α≥1, the throughput of this node will suddenly decrease[9]. However, the capacity of the entire network will not keep increasing along with the addition of the nodes. More badly, the capacity probably decreases and all the nodes collide. On the other hand, the increase of the covered area not only means the number of nodes increases, but also leads to the decrease of system throughput because of the reduce of average data transmission chance.
The nearer to the central transfer node a sensor node is, the heavier its data transfer traffic is. Therefore, the sensor nodes near to the central transfer node are easy to become a bottleneck for system capacity and reliability. For the WSN, a key collision area exists around the central transfer node. The nodes in this area have to contend for channels for the traffic they send or transfer. The total traffic under the contention is approximately in positive proportion to the total traffic originally sent by all the nodes of the system. Under the condition that the traffic originally sent by every node is the same and that the fair access is guaranteed, the throughput each node can get is just 0(1/n).
Parallel transmission technology improves the capability of the sensors of the central transfer node for parallel transmission and frequency multiplexing. Therefore, it can be used to improve the system capacity of the WSN, and to reduce the ratio of collision and congestion. Moreover, the advanced technologies of adaptive power control, interference cancellation (multi-user detection) and interference utilization (diversity reception and MIMO) may be used for adjacent nodes to have more than one channel to use. Multiple channels make the original contention simple. This belongs to the resource scheduling. Access delay is the key factor in the performance of a WSN. Data transfer among multiple nodes means a relatively long processing time delay. The access delay is determined by the transfer processing time delay of nodes and the time delay at the link layer for contention and queuing for retransfer. The former is related to the distribution of nodes and network coverage, while the later is effected by average nodes contending with the node and the traffic of the nodes in contention. The WSN should not fail to use the right adaptive policies for links (such