Optimal Multi-Sensor Deployment via Sample-Based Quality-of-Service Distribution Matching

This paper considers a multi-sensor service matching deployment problem over a set of discrete target points that populate a finite flat surface. The service can be event detection among targets using a vision sensor or an acoustic receiver, video surveillance for target monitoring, or providing wireless coverage to the targets. The quality-of-service (QoS) of the sensors is spatially nonuniform and can be anisotropic. The sensors are heterogeneous in the sense that their QoS distribution over their sensing footprint is not the same. The objective is to determine the sensor’s best deployment position and orientation such that the collective multi-sensor QoS distribution matches the spread of the targets in the environment as closely as possible. To solve this problem, we propose a two-stage deployment strategy. First, we partition the environment using the computationally efficient K-means clustering algorithm. Then, we sample points from the QoS distribution over the sensing footprint. Then, for each sensor-cluster pair, we use an iterative closest-point approach inspired by the point cloud registration algorithms used in computer vision to determine the best deployment position and orientation for the sensor. Finally, we use a linear assignment problem framework to assign the clusters to the sensors. Numerical examples demonstrate our results.