Trustworthy Traffic Sensors Networks via Fault-Tolerant Sensor Reconciliation Architectures

In this paper, the sensor selection problem (i.e. the problem of selecting a subset of $p\ll n$ possibly redundant sensors from a larger set of $n$ potential sensors) for large-scale Wireless Sensor Networks (WSNs) is considered along with the concomitant design problem of determining a suitable fault-tolerant full state observer for effective state reconstruction in the presence of sensor faults/failures. Given a set of redundant physical sensors, a monitoring architecture for estimating traffic flows via a novel fault-tolerant sensor reconciliation design procedure is proposed that is able to hide measurements that may be corrupted by sensor errors and provide reliable state estimates.The proposed approach utilizes a Linear Parameter Varying (LPV) Luenberger observer, which is responsible for robust and reliable system state reconstruction. The results show that the proposed estimation strategy has superior performance in both minimizing state reconstruction errors and estimating bias errors and sensor effectiveness when compared to conventional strategies and with another state-of-the-art approach exploiting Takagi-Sugeno (TS) fuzzy observers. Finally, the effectiveness of the reconciliation architecture has been evaluated in realistic road infrastructure and traffic data provided by the software Aimsun Next.