Abstract:We consider the problem of sensor selection for event detection in wireless sensor networks (WSNs). We want to choose a subset of p out of n sensors that yields the best detection performance. As the sensor selection optimality criteria, we propose the Kullback-Leibler and Chernoff distances between the distributions of the selected measurements under the two hypothesis. We formulate the maxmin robust sensor selection problem to cope with the uncertainties in distribution means. We prove that the sensor selection problem is NP hard, for both Kullback-Leibler and Chernoff criteria. To (sub)optimally solve the sensor selection problem, we propose an algorithm of affordable complexity. Extensive numerical simulations on moderate size problem instances (when the optimum by exhaustive search is feasible to compute) demonstrate the algorithm's near optimality in a very large portion of problem instances. For larger problems, extensive simulations demonstrate that our algorithm outperforms random searches, once an upper bound on computational time is set. We corroborate numerically the validity of the Kullback-Leibler and Chernoff sensor selection criteria, by showing that they lead to sensor selections nearly optimal both in the Neyman-Pearson and Bayes sense.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is the sensor selection problem in event detection in wireless sensor networks (WSNs). Specifically, the researchers hope to select p sensors (p < n) from n sensors to achieve the best detection performance. Here, "best" means that under the two hypotheses (event occurs or does not occur), the distance between the distributions of the measured values of the selected sensors is the largest, thereby improving the detection accuracy.
### Problem Background
In wireless sensor networks, due to limited power and communication resources, how to balance the operating cost and the quality of observation is an important challenge. The core of the sensor selection problem lies in how to select a subset of sensors that can provide the most valuable information while limiting the operating cost.
### Specific Problem Description
- **Hypothesis Setting**: The natural state can have two situations: \( H_1 \) (event occurs, for example, the target exists) and \( H_0 \) (event does not occur, for example, the target does not exist).
- **Sensor Network**: A wireless sensor network composed of n sensors is used to monitor the natural state.
- **Distribution Assumption**: It is assumed that under the two hypotheses, the distribution of the n - dimensional measurement vector is a Gaussian distribution, with different means \( m_i \) and covariance matrices \( S_i \), where \( i = 0, 1 \).
- **Selection Constraint**: Due to the inherent constraints of wireless sensor networks (such as power limitations), only p sensors can perform measurements and transmit data to the fusion node. The fusion node performs detection based on the received p readings, that is, decides which hypothesis is correct.
- **Target Problem**: Which p sensors should be selected to achieve the best detection performance?
### Optimization Criteria
To evaluate the optimality of sensor selection, the author proposes two criteria:
1. **Kullback - Leibler (KL) Distance**: Measures the difference between two distributions.
2. **Chernoff Distance**: Measures the difference between two distributions and is more robust especially in uncertain situations.
### Mathematical Representation
Given the density functions \( f_1 \) and \( f_0 \) of two distributions, the KL distance and the Chernoff distance are defined as follows respectively:
\[ D_{\text{KL}}(f_1 \| f_0) := \int \log \frac{f_1(x)}{f_0(x)} f_1(x) \, dx \]
\[ D_{\text{C}}(f_1, f_0) := \max_{s \in [0,1]} -\log \int f_1^s(x) f_0^{1 - s}(x) \, dx \]
### NP - Hardness of the Problem
The author proves that the sensor selection problem using the KL distance and the Chernoff distance as optimization criteria is NP - hard, which means that it is difficult to find an efficient algorithm to solve large - scale instances.
### Solution
To (sub -) optimally solve the sensor selection problem, the author proposes an algorithm with lower computational complexity. The algorithm is divided into three stages:
1. **Relaxation Stage**: Relax the problem from a combinatorial optimization problem to an optimization problem of the Stiefel matrix.
2. **Projection Stage**: Project the Stiefel matrix back to the 0/1 selection matrix.
3. **Refinement Stage**: Perform local optimization in the neighborhood of the 0/1 selection matrix to further improve the performance.
### Experimental Results
Through extensive numerical simulations, the author shows that the algorithm is close to optimal in most cases and outperforms the random search method in large - scale problems. In addition, the experiments also verify the effectiveness of the KL distance and the Chernoff distance as optimization criteria, especially performing well in the detection performance in the Neyman - Pearson and Bayes senses.
In summary, this paper aims to improve the event detection performance in wireless sensor networks by optimizing sensor selection, proposes optimization criteria based on the KL distance and the Chernoff distance, and designs an effective algorithm to solve this NP - hard problem.