Within the POMDP framework, the action space is usually infinite and continuous, but in practice, it is often assumed to be a finite set of actions.ĭifferent methods have been proposed to solve the optimal policy problem.

In this way, a decision-theoretic approach can be taken, leveraging the sensor nodes observations and the reward from following management actions to customize choices. POMDP allows the calculation of the optimal policy for choosing the desired actions despite some important information may not be observed. The sensor selection can be framed as a linear programming problem based on the partially observable Markov decision process (POMDP) framework. In this paper, we consider the optimal selection of the subset, adapting to the position of the target in an online manner. These two challenges motivate the problem of selecting only a small subset of the sensors for the measurement purpose.

On the other hand, not all the nodes are used simultaneously due to (a) performance requirements (e.g., a small number of sensors is sufficient to achieve a desired tracking accuracy) and (b) the constrained computation and communication resources (e.g., only a small number of sensors can be activated at the same time). On the one hand, the maneuvering target moves across the FoVs of different sensors, for which the target trajectory cannot be covered by the FoV of any single sensor. However, the limited power and wireless-bandwidth impose challenges. The neighbor sensors have overlap field of views (FoVs) so that they can cooperate in the common FoV for better tracking accuracy and system robustness. In general, the sensors have limited computational resources and communication range, covering a small part of the region of interest (ROI).

Countless applications rely on wireless sensor networks of geographically dispersed sensors, e.g., target tracking, environment monitoring and radar network resource management.

Network radar crlb