Clustering and sensing with decentralized detection in vehicular ad hoc networks

In the near future, vehicles will be more and more advanced sensing platforms: for instance, at least one smartphone (with several on-board sensors) is likely to be inside each vehicle. Smartphone-based inter-vehicle communications thus support the creation of vehicular sensor networks (VSNs). In this paper, we analyze the performance of clustered VSNs, where (hierarchical) decentralized detection schemes are used to estimate the status of an observed spatially constant phenomenon of interest. Clustering makes processing efficient and the architecture scalable. Our approach consists of the creation, during a downlink phase, of a clustered VSN topology through fast broadcast of control messages, started from a remote sink (e.g., in the cloud), through a novel clustering protocol, denoted as cluster-head election irresponsible forwarding (CEIF). This clustered VSN topology is then exploited, during an uplink phase, to collect sensed data from the vehicles and perform distributed detection. The performance of the proposed scheme is investigated considering mostly IEEE 802.11b (smartphone-based) as well as IEEE 802.11p (inter-vehicle) communications in both highway-like and urban-like scenarios. Our results highlight the existing trade-off between decision delay and energy efficiency. The proposed VSN-based distributed detection schemes have to cope with the “ephemeral” nature of clusters. Therefore, proper cluster maintenance strategies are needed to prolong the cluster lifetime and, as a consequence, the maximum amount of data which can be collected before clusters break. This leads to the concept of decentralized detection “on the move.”