Sentinel surveillance of traffic conditions with multilayer network

Surveillance of traffic condition by sensors is widely used in society to monitor the traffic condition of roads. To save budget and lengthen sensors’ lifetime, it is thus of paramount importance to estimate the dynamical traffic condition of all the targeted locations by partial observing a minimum number of sensors. As thus, we do not need to deploy or open all of the expensive sensors of traffic condition everywhere and every time in the process of monitoring. Previous studies focus on the finding of a group of sentinel sensors with a single objective of precise estimation. However, the traffic sensors located in the city are inevitably influenced by their social/natural environment (e.g., social management and power) with complex interactions. In this paper, we propose a dynamic optimization model of sensor location selection for the sentinel surveillance of dynamical traffic condition under the realistic social and natural environment of a city. We use some non-concave items to model the interactions among sensors, which are the challenge to infer via nonlinear optimization algorithms. As thus, we give the details of alternating direction method of multipliers algorithm for our model, with the capability to deal with large interactions of variables. Taking the traffic system of Shanghai as a case study, our experiments show the proposed model’s performance both in accuracy and flexibility of different balance among precision and natural/social constraints.