In order to solve the problem of selecting the optimal combination of sensors and determining their optimal placement in a surveillance system, Sivaram, Kankanhalli, and Ramakrishnan design an integer programming model and explore some characteristics related to nonconvexity.
The paper is divided into six sections. The first is an introduction, where the problem is well defined and established. Sections 2 and 3 present related work for the problem and the proposed technique. The technique consists of a performance vector whose elements represent the performance of subtasks, and a performance matrix that is presented with this vector. The performance vector corresponds to a sensor combination. As the problem is combinatorial, the integer linear model is useful for finding the optimal combination.
Section 4 is devoted to the design of the surveillance system that consists of two types of sensors: infrared cameras and active motion sensors. The effect of the sensors and the cameras on the image capture subtask is modeled. This is an important section; it is very clear, but lacks a brief description of the combinatorial problem involved. Section 5 presents the results, but it doesn’t explain the difficulty of working with nonconvex problems or offer statistics related to different experiments with the problem. The conclusions in the last section show the effectiveness of the technique based on the experimental results.
Overall, the paper is clear and well organized. However, it would have been stronger if it had described the nonconvex nature of the problem and the computational complexity of the combinatorial problem, and given more details on the experiments.