Distributed Sensor Fusion Based on Statistical Inference.
Abstract
We addressed several issues related to distributed detection and estimation problems. We examined the posterior robustness of decentralized detection schemes when the prior density is not completely known. The results indicate the degree of robustness that could be achieved with quantized observations. We developed a new CFAR test using distributed sensors for detecting radar targets. The numerical results for Rayleigh target in Rayleigh clutter indicate that, Under certain assumptions, the newly developed MOS test performs considerably better than the OS-CFAR detector with the AMC For the OR fusion rule. As shown here, the performance loss associated with a distributed detection scheme can be assessed using an easily computable probability of error expression, when the density of sensor observation belongs to the exponential family. In distributed parameter estimation we numerically evaluated the mean squared error performances of several estimation schemes for two example situations. Finally, we have analyzed a rank based test for M-ary detection problems. The results show that the new modified rank test can provide performance gain in detecting signals in non-Gaussian noise.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1997
- Accession Number
- ADA327114
Entities
People
- Ramanarayanan Viswanathan
Organizations
- Southern Illinois University Carbondale