Wire generates and maintains one consistent world state estimate based on object detections. It solves the data association problem by maintaining multiple hypotheses and facilitates tracking of various object attributes. The state estimators used for estimation and the probabilistic models used for association can be configured. Technical details can be found in this paper:
J. Elfring, S. van den Dries, M.J.G. van de Molengraft, M. Steinbuch, Semantic world modeling using probabilistic multiple hypothesis anchoring, Robotics and Autonomous Systems, Volume 61, Issue 2, February 2013, Pages 95-105, (pdf) which also includes a more detailed explanation of the algorithm.
Tutorials can be found on the ROS wiki.