1: \begin{abstract}

2: The ability of a sensor node to determine its physical location within

3: a network (Localization) is of fundamental importance in sensor

4: networks. Interpretating data from sensors will not be possible unless

5: the context of the data is known; this is most often accomplished by

6: tracking its physical location.  Existing research has focused on

7: localization in static sensor networks where localization is a

8: one-time (or low frequency) activity.  In contrast, this paper

9: considers localization for mobile sensors: when sensors are mobile,

10: localization must be invoked periodically to enable the sensors to

11: track their location.  The higher the frequency of localization, the

12: lower the error introduced because of mobility.  However, localization

13: is a costly operation since it involves both communication and

14: computation. In this paper, we propose and investigate adaptive and

15: predictive protocols that control the frequency of localization based

16: on sensor mobility behavior to reduce the energy requirements for

17: localization while bounding the localization error.  We show that such

18: protocols can significantly reduce the localization energy without

19: sacrificing accuracy (in fact, improving accuracy for most

20: situations).  Using simulation and analysis we explore the tradeoff

21: between energy efficiency and localization error due to mobility for

22: several protocols.

23: \end{abstract}

24: