A Purpose-Oriented 3-D Voronoi Algorithm for Deployment of a Multitype Sensor Network in Complex 3-D Indoor Environments in Support of the Mobility of People With Motor Disabilities

Sensor networks are increasingly used for creating smart interactive environments to support a wide range of applications, including mobility, transportation, security, and emergency evacuation. Multitype sensor networks, in particular, have the potential to assist people with motor disabilities (PWMDs) by providing them with essential accessibility information during their mobility in complex indoor environments. Nevertheless, the optimal placement of such a sensor network is still a very challenging task. Several solutions have been proposed to maximize the spatial coverage of multitype sensor networks in recent years. However, most of these solutions have limitations when it comes to deploying a multitype sensor network in a complex indoor environment adapted to support the mobility of PWMD. This is due to the complexity of 3-D indoor environments and the specific needs to support the mobility tasks of potential to assist people with motor disability (PWMD) (e.g., priority definition for the coverage areas, consideration of obstacles, and facilitators present in the environment). This article proposes a purpose-oriented 3-D Voronoi (PO-3DVOR) algorithm for the optimal deployment of a multitype sensor network in a 3-D complex indoor environment in support of the mobility of PWMD. For this purpose, we propose an integrated algorithm using IndoorGML and a 3-D Voronoi spatial data structure for the management of information from the indoor environment and the multitype sensor network, respectively. We also integrate specific information about the complexity and legibility of the environment for PWMD to guide sensor network optimization and maximize the purpose-oriented weighted coverage assessment of the sensor network. To test and validate the proposed algorithm, we consider two types of sensors for implementation: cameras (long-range sensors) and Bluetooth sensors (short-range sensors). We then compare the proposed local optimization algorithm with the covariance matrix adaptation evolution strategy (CMA-ES), which is a global optimization approach. Experimental results demonstrate that the proposed algorithm achieves better performance compared to the CMA-ES algorithm in terms of computing time and coverage.