Time Synchronization for Wireless Sensor Networks - Adaptive Signal Processing in Wireless Communications - page 386

applied to RRS. Consequently, there is no difference between ROS and RRS with regard
to the accuracy of synchronization since the effects of random delays are the same. Like-
wise, the CRB for RRS can also be obtained using a similar procedure as in ROS. When
there is no relative clock skew (θ (BA) = 0), it is straightforward to show that the maximum
likelihood estimator of the relative clock offset θ ˆ (BA) becomes

N

1

∑

ˆ


 


 

θ (BA)

() A () B

= TT − , (13. 29)

2 , i 2 , i

N

i = 1

which is the equivalent to the estimator presented in [18].
The main benefit of this approach is that all nondeterministic delay components on
the transmitter side (send time and access time) are eliminated. Thus, a high degree of
synchronization accuracy can be achieved using this approach.

13.4 Existing Time Synchronization Protocols

Thus far, a number of protocols have been suggested to solve the problem of time syn-
chronization in distributed networks. For general computer networks, NTP has been
adopted as the standard time synchronization scheme of the Internet [3]. Although NTP
was shown to perform well in computer networks, it is not directly applicable to wire-
less sensor networks due to the unique challenges sensor networks face: limited power
resources, wireless channel conditions, dynamic topology changes, etc. (recall also the
design considerations presented in section 13.2.2). NTP enjoys unlimited (or recharge-
able) energy resources and a relatively static topology in computer networks. However,
these are not available in sensor networks. Therefore, different types of time synchroni-
zation protocols have been proposed to meet the design requirements of wireless sensor
networks [2].
Ideally, a time synchronization protocol should be able to work optimally in terms
of all design requirements of time synchronization, which are energy efficiency, scal-
ability, precision, security, reliability, and robustness to network dynamics. However,
the complex nature of wireless sensor networks makes it very difficult to optimize the
protocol with respect to all these requirements simultaneously. Due to the trade-offs
in satisfying these requirements, each protocol is designed to put distinct emphases on
different requirements.
Assuming various criteria, time synchronization protocols can be categorized into
different classes:

• Master-slave versus peer-to-peer

• Master-slave : Where first a tree-like network hierarchy is arranged, and

upon the completion of this arrangement only the connected nodes in the
hierarchy synchronize with each other.
Peer-to-peer

• : Where any pair of nodes in the network can synchronize

with each other.