Many-Objective Sensor Selection in IoT Systems
The Internet of Things connects physical objects through sensor devices with multiple functionalities. At the planning stage of deploying an IoT system, we are concerned about sensor selection in the IoT system, which allocates predefined IoT services to multiple sensor devices so as to optimize one or more objectives associated with these allocations, under energy and distance constraints. The sensor selection problem that optimizes a utility function in other applications has been shown to be NP-hard, and the number of IoT services concerned is enormous in practice. Hence, it is suitable to apply evolutionary algorithms (EAs) for solving the large-scale problem with multiple objectives. Recently, the paradigm of multiple-objective EAs (which often address only two or three objectives) has advanced to many-objective EAs (which are intended to address four or more objectives that may be in conflict with each other in many cases). Therefore, this article considers many objectives of the sensor selection problem in the IoT system, including optimization of communication energy consumption, energy balancing on all devices, energy harvesting, green concerns, and QoS. The problem is resolved by a tailored many-objective EA based on decomposition to increase computational efficiency and solution quality. By simulation, the proposed EA is shown to be promising through scatter charts and parallel coordinates.
Most of the sensor selection problems covered in previous works considered only one or two objectives to be optimized. In practice, however, sensor selection often involves many objectives, some of which may be in conflict with each other in many cases, and is required to satisfy a lot of constraints. Recently, a lot of research has successfully extended multi-objective optimization problems (MOPs) to many-objective optimization problems (MaOPs), in which the former often focus on only two or three objectives, whereas the latter generally consider optimization of four or more objectives. For the MaOP, with increased number of objectives and the solution search space, a large number of non-dominated solutions lead to selection pressure on the EA, and it is hard to achieve convergence and diversity of the EA. In addition, it is hard to estimate the amount of computing resources required to solve MaOPs.
Shifted to MaEAs for MaOPs. Most of the MaEAs improve conventional MEAs by some advanced techniques in evolutionary computation. Among major MOEAs for MOPs , the MOEA based on decomposition (MOEA/D)  is one of the most popular MOEAs. The MOEA/D is often extended to solve MaOPs. For instance, the work in  based on a normal boundary intersection method to decompose an MaOP into multiple subproblems, and then proposed an improved MOEA/D for the problem, which applies a systematic sampling scheme to produce uniformly distributed reference points, and applies two independent distance measures to maintain the balance between convergence and diversity. The work in  proposed an MOEA/D with sorting and selection for MaOPs, which associates various solutions with the same subproblems, and flexibly allows some subproblems with no associated solutions. The work in  proposed an MOEA/D with localized weighted sum for MaOPs, which selects each optimal solution search direction only from the neighboring solutions of the current solution.
The recent advances in EA have shifted from the paradigm of multi-objective optimization to many-objective optimization. Therefore, this article provides a comprehensive review of various sensor selection problems in IoT systems, and proposes a sensor selection problem with five objectives in IoT systems. We then solve the problem by an MOEA/D. The major contributions of this work are summarized as follows: 1. This work is the first to propose the MaOP for sensor selection in IoT systems. 2. An MOEA/D approach is proposed for addressing the problem. 3. To evaluate performance of the proposed approach, detailed experimental analysis is conducted through box plots, scatter charts, and parallel coordinates. Box plots of simulation results show that increase of the problem size leads to increase of energy consumption, energy balancing, number of EH services, and pollution level, while it does not affect QoS remarkably. Scatter charts and parallel coordinates of simulation results show that the lower energy balancing, the lower energy consumption, the lower QoS, the higher pollution level, and more EH services.
 L. D. Xu, W. He, and S. Li, “Internet of Things in Industries: A Survey,” IEEE Trans. Industrial Informatics, vol. 10, no. 4, Nov. 2014, pp. 2233–43.
 Z. Huang et al., “Communication Energy Aware Sensor Selection in IoT Systems,” Proc. IEEE Int’l. Conf. Internet of Things,d Green Computing and Communications, and Cyber, Physical and Social Computing, Sept. 2014, pp. 235–42.
 Z. Huang, K. J. Lin, and L. Han, “An Energy Sentient Methodology for Sensor Mapping and Selection in IoT Systems,” Proc. 2014 IEEE 23rd Int’l. Symp. Industrial Electronics, June 2014, pp. 1436–41.
 F. Bian, D. Kempe, and R. Govindan, “Utility Based Sensor Selection,” Proc. 5th Int’l. Conf. Info. Processing Sensor Networks, ACM Press, Apr. 2006, pp. 11–18.
 S. Joshi, and S. Boyd, “Sensor Selection via Convex Optimization,” IEEE Trans. Signal Processing, vol. 57, no. 2, Feb. 2009, pp. 451–62.
 S. Y. Yu et al., “QoS Oriented Sensor Selection in IoT System,” Proc. IEEE Int’l. Conf. Internet of Things, Green Computing and Communications, and Cyber, Physical and Social Computing, Sept. 2014, pp. 201–06.
 T. Xu and M. Potkonjak, “Energy Saving Using Scenario Based Sensor Selection on Medical Shoes,” Proc. 2015 Int’l. Conf. Healthcare Informatics, IEEE Press, Oct. 2015, pp. 398–403.
 N. H. Viet, E. Munthe-Kaas, and T. Plagemann, “Energy-Balanced Sensor Selection for Social Context Detection,” Proc. 2012 IEEE Int’l. Conf. Pervasive Computing and Commun. Wksps., IEEE Press, Mar. 2012, pp. 32–37.
 T. Yan et al., “An Energy-Efficient Sensor Selection Strategy Based on Spatial Correlation for Wireless Sensor Networks,” Proc. 2014 6th Int’l. Conf. Wireless Commun. and Signal Processing, IEEE Press, Oct. 2014, pp. 1–6.
 M. Calvo-Fullana, J. Matamoros, and C. Ant, “Sensor Selection in Energy Harvesting Wireless Sensor Networks,” Proc. 2015 IEEE Global Conf. Signal and Info. Processing, Dec. 2015, pp. 43–47.