OS-CFAR and CMLD threshold optimization in distributed systems using evolutionary strategies

Abstract

This paper proposes an improvement of the threshold optimization in distributed ordered statistics constant false alarm rate and censored mean level detector using Evolutionary Strategies (ESs). The target is assumed to be Rayleigh distributed and the observations are independent from sensor to sensor. Two fusion rules; "AND" and "OR" were considered. An ES was tested and a comparison with a genetic algorithm improved by a tournament selection was also analyzed. Among a variety of evolution strategies, the most popular proposed in the literature are the strategy (µ, λ) and the strategy (µ + λ). We proposed an (µ + λ) evolution strategy, by which a self-adaptation mutation is used. The results showed that, although the ES is more difficult to implement and is in a certain manner slower than the GA, it improves the performance of the system.

Robust LFT-LPV H∞ Control of an Underactuated Inverted Pendulum on a Cart with Optimal Weighting Functions Selection by GA and ES

“…In the simulation, we don't apply the robust LPV controller on the approximated LPV model ( 6)- (7), but on the original nonlinear model itself (3). Fig.…”

Performance comparison of mean-level CFAR detectors in homogeneous and non-homogeneous Weibull clutter for MIMO radars

“…The first category includes the mean-level CFAR detectors, where the sum of reference cells is used to estimate the noise power level; the cell averaging (CA-CFAR) proposed by Finn and Johnson [5] and its variants; the greatest of (GO-CFAR) and the smallest of (SO-CFAR) [6,7]. The second category is based on the ordered statistic technique, which includes the ordered statistic CFAR (OS-CFAR) [8,9] and its variants; the generalized ordered statistics cell averaging CFAR (GOSCA-CFAR), the ordered statistic smallest of (OSSO-CFAR), the ordered statistic greatest of (OSGO-CFAR) and the generalized switching CFAR (GS-CFAR) [4,10,11]. In the third category, in order to exploit the advantages of the two techniques, the noise power level estimation uses a combination of the sum of the reference cells and the order statistic technique; the mean of OS and CA-CFAR (MOSCA-CFAR), the smallest of OS and CA-CFAR (SOSCA-CFAR), the OS and CA-CFAR greatest of (OSCAGO-CFAR) and the generalized SOSCA-CFAR [12].…”