Abstract
•The data fusion problem is considered over wireless sensor networks.•The CO2 concentration is monitored in a greenhouse.•The consensus-Kalman filter is designed for the distributed estimation scheme.•The packet loss phenomenon is reflected by the proportion of the faulty sensors.•The newly developed filtering algorithm is suitable for online applications.
This paper investigates the data fusion problem over wireless sensor networks (WSNs) for monitoring the carbon dioxide (CO2) concentration in a greenhouse. CO2 concentration is an important environmental parameter in the greenhouse, and adequate regulation of CO2 concentration is certainly beneficial to the improvement of the crop growth efficiency in the greenhouse. Since the measurement of CO2 concentration is unavoidably subject to environmental interferences (e.g. noises), it is vitally important to estimate the true CO2 concentration through available sensor measurements over a WSN with given topology. In this paper, based on the Consensus–Kalman filter, the distributed estimation scheme is presented to improve the state estimation accuracy. After the distributed consensus estimation, the head node can conduct the fusion estimation on the data from the available sensor nodes. The packet loss phenomenon brought by unreliable communication links is reflected by the proportion of the faulty sensors. To further alleviate the effects from the packet losses, we propose a modified estimation scheme for the head node that combines the estimates from sensor node at both the previous and current time points, thereby enhancing the accuracy of data fusion. Simulation analysis is carried out on the collected information of CO2 concentration in the greenhouse in order to demonstrate the effectiveness of the proposed scheme.