Reliable Traffic Sensor Networks Via Fault-Tolerant Sensor Reconciliation Schemes

Abstract

Wireless sensor networks (WSNs) are in general composed of a large number of sensor nodes spread over a wide geographical area. In order to minimize the processing/communication burdens and hardware/maintenance costs, a limited number of sensors must reasonably be used and they arguably need to be smartly located. Moreover, unavoidable failures may affect the deployed sensors and may cause wrong elaborations and errors in the system state estimates, degrading the system performance or causing instability. This paper presents a traffic flow estimation architecture based on a fault-tolerant reconciliation design procedure for over-sensed sensor networks. In particular, given a subset of p redundant physical sensors, determined from the outset form a larger set of n ≫ p potential sensors through the solution of an Observability-based sensor selection procedure, the goal of the reconciliation observer is to hide corrupted measures in the estimation process, possibly generated by faulty sensors, and provide reliable state estimates. The presented approach envisages the use of Linear Parameter Varying (LPV) Luenberger Observers in charge of robustly estimating the system state along with the bias and the loss of effectiveness sensor faults. A road traffic monitoring problem is used as a case study to demonstrate the effectiveness of the proposed strategy.