Trunk and Trail: Wireless Sensor Network Services For Distributed Object Tracking

In this paper we design two wireless sensor network services Trunk and Trail, for distributed, mobile object tracking applications. Both these services are designed to be reliable and energy efficient while satisfying their specification. The specification for Trunk is to return a consistent global snapshot of the state of all objects in the system to all subscribed applications at regular intervals of time. Because the queries are specified apriori and are periodic Trunk operates in a synchronous model, and therefore the nodes in the network listen on radio and transmit only when scheduled, thus being energy efficient. The specification for Trail is to return the location of a particular mobile object to any application issuing the query. Trail maintains a tracking data structure for all objects and this structure is updated locally upon object moves and using this structure Trail offers a find time that increases linearly with the distance from an object. However Trail operates in an asynchronous model where the different application queries are not synchronized and nodes in the network are always awake to listen queries. Since this is an energy consuming, we design a Synchronous Trail service in which the algorithms for maintaining the tracking data structure and the find operations are like in Trail, but the nodes in the network operate synchronously and thereby gaining energy efficiency at the cost of latency. We then consider an example mobile object tracking system, called an intruder-interceptor game and describe how the interceptor application uses our network services Trunk and Trail. We also provide an experimental analysis of the performance of these two services in terms of their reliability and latency in a network of 105 XSMs (eXtreme Scale Motes) using Kansei, a wireless sensor network testbed.