Mobile Sensing of Multi-Dimensional Dynamic Field via Compressed Sensing

In this paper, we consider to measure time-varying dynamic signals at discrete locations by using a single mobile sensor The challenge arises as the mobile sensor is required to transit between the sampling locations, resulting in intermittent measurements at each location, and the time-varying dynamic signals must be recovered from the intermittent measured data. In this work, we propose to tackle this single mobile sensing of multi-location dynamic signals through the compressed sensing framework. Both the temporal-spatial coupling arising from the random sampling requirement and the mobility limitation of the sensor in transition between sampling locations are addressed through a simulated annealing based optimization approach. Simulation results are presented and discussed to illustrate the proposed method.

Two equivalent multi-sensor Kalman filters with variable delays and intermittent measurements

Kalman filtering of measurement data from multiple sensors with time-varying delays and missing measurements is considered in this work. Two existing approaches to Kalman filtering with delays are extended by removing some assumptions in order to have equivalent filtering methods and making comparisons between them. The computational loads of the two methods are compared in terms of the average number of floating point operations required by each method for different system dimensionalities and delay upper bounds. The results show that the superiority of the methods over each other depends on the comparison conditions.

Two equivalent multi-sensor Kalman filters with variable delays and intermittent measurements

Kalman filtering of measurement data from multiple sensors with time-varying delays and missing measurements is considered in this work. Two existing approaches to Kalman filtering with delays are extended by removing some assumptions in order to have equivalent filtering methods and making comparisons between them. The computational loads of the two methods are compared in terms of the average number of floating point operations required by each method for different system dimensionalities and delay upper bounds. The results show that the superiority of the methods over each other depends on the comparison conditions.