Wireless Sensor and Actuator Networks

32 Fundamentals of WSANs design

A quick diagnosis of conditions will be realized. In case a relevant parameter on a

machine is approaching a critical threshold, Ada will be alerted on her device by an alert

message. For example, sensors could detect that the fruit press machine overheats and

that the level of lubrication oil is abnormally low. An alert is sent to Ada with the infor-

mation about the thresholds and the location of the machine.

2.2 Event detection and spatial and time random

process estimation

With the aim of creating a very general taxonomy of WSN applications, they could be

classifi ed into the following two categories: event detection (ED) and spatial and time

random process estimation (PE).

In the fi rst type of application, sensors must detect an event, for example a fi re in a for-

est, a quake, etc. Therefore, no periodic monitoring by the sink(s) is required. Rather, the

nodes must be able to detect an event that might happen at any time in any location within

the monitored space.

In the second case, the WSN aims at periodically estimating a given physical phenomenon

(e.g., the atmospheric pressure in a wide area or the ground temperature variations in a

small volcanic site) which can be modelled as a bi-dimensional random process (generally

non-stationary). If we consider a typical application related to data gathering from an area

and forwarding data to a fi nal sink(s), when sensors receive a service request they take a

sample from the environment and transmit it by following an appropriate communication

protocol to the fi nal sink(s), which is in charge of collecting all information detected by

nodes and estimating the process realization.

The two types of applications are characterized by separate user requirements and

impose specifi c characteristics at the WSN.

In the ED applications the signal processing within devices is very simple, owing to

the fact that each device has to compare measurement results to a given threshold and

send the binary information to the fi nal sink(s). The density of nodes must ensure that

the event is detected with given probability; therefore the network must have a suffi -

cient coverage. The network coverage is related to the sensing range of nodes and the

event type. Moreover, distributed localization algorithms could be used to identify the

exact position in which the event occurs. The report can be received by the sink(s) with

given probability; connectivity issues, related to transmission range of nodes, must be

kept under control and the communication protocols must be designed so that the alarm

arrives at the sink(s) with a high probability and in a short time.

Applications of WSANs 33

In case the sink(s) periodically queries the nodes, the frequency of queries must be

chosen so that the event is detected with a given probability and must also ensure that

the reports reach the fi nal sink(s) in time.

Therefore, the main issues which characterize this type of application are:

●

coverage

●

distributed localization

●

connectivity

●

communication protocols (they must ensure low losses of packets and low delays).

Typical requirements for ED applications are therefore:

●

minimum probability of coverage

●

maximum localization error

●

minimum probability of connectivity

●

maximum packet loss probability

●

maximum packet delivery delay

●

and so on.

As far as PE applications are concerned, the density of nodes must ensure that the pro-

cess is accurately estimated; therefore signal processing is one of the main issues. The

samples can be received by the sink(s) with given probability; connectivity of the net-

work, related to transmission range of nodes, must be kept under control and the com-

munication protocols must be designed so that the process estimation error is maintained

under a given threshold. Here the sampling frequency must be chosen so that the process

evolution is tracked.

Therefore the main issues which characterize this type of application are:

●

signal processing

●

connectivity

●

time synchronization

●

communication protocols (they must ensure low process estimation error).

Typical requirements for PE applications are therefore:

●

maximum estimation error

●

maximum localization error

●

minimum probability of connectivity

●

maximum packet loss

●

and so on.