A Hierarchy-Based Fault-Local Stabilizing Algorithm for Tracking in Sensor Networks

In this paper, we introduce the concept of hierarchy-based fault-local stabilization and a novel self-healing/fault-containment technique and apply them in Stalk. Stalk is an algorithm for tracking in sensor networks that maintains a data structure on top of an underlying hierarchical partitioning of the network. Starting from an arbitrarily corrupted state, Stalk satisfies its specification within time and communication cost proportional to the size of the faulty region, defined in terms of levels of the hierarchy where faults have occurred. This local stabilization is achieved by slowing propagation of information as the levels of the hierarchy underlying Stalk increase, enabling more recent information propagated by lower levels to override misinformation at higher levels before the misinformation is propagated more than a constant number of levels. In addition, this stabilization is achieved without reducing the efficiency or availability of the data structure when faults don’t occur: 1) Operations to find the mobile object distance d away take O(d) time and communication to complete, 2) Updates to the tracking structure after the object has moved a total of d distance take O(d ∗ log network diameter) amortized time and communication to complete, 3) The tracked object may relocate without waiting for Stalk to complete updates resulting from prior moves, and 4) The mobile object can move while a find is in progress.