소나시스템에서 유전알고리듬을 이용한 최적음향탐색경로전략

Abstract

When a searcher detects the target using sonar in complicated ocean environments, the calculation of the optimal sonar search track is an important influence on the effectiveness of sonar and human resources. In addition, because the ladder search method in general use is intuitively not the optimal search method, the development of a search path planning method with improved performance and reduced search time is an important research focus.

The optimal search path problem can be treated as a search-effort-allocation problem, which assumes that the effort can be allocated arbitrarily over the search space within the achievable path by the sonar platform. The search path can be modeled by either the discrete-search-path problem, which assumes that the searcher and target move in discrete space and time, or the continuous-search-path problem, which assumes that they must follow realizable paths in continuous space and time. Recently, DelBalzo[1] developed a calculation method for the continuous-search-path problem based on the combination of the genetic algorithm(GA) and the detection range.

In this study, GA is used in non-homogeneous and anisotropic environments to nearly optimize the sonar search track[6], and Bayesian statistics allow amalgamation of the individual detection probability into a Cumulative Detection Probability (CDP, ) for the search path[5], which is the Measure Of Effectiveness (MOE) for that path against a distribution of targets[4]. The optimization metric for the search path is the target CDP during a fixed time period[6].

As for the path, the discrete-search-path and the continuous-search-path is employed, but the search step is fixed in length and time. The movement direction of the searcher is used as the gene of GA, which means that each gene is only composed of one set of real numbers, , representing the direction of movement, so that . In addition, due to the process of evolution the offspring of each generation contain a wide variety of candidate paths for perturbing some aspects of the trial solution. Crossover is accomplished by exchanging the genes between the initial and final segments of the two parents. The perturbation and elimination of the nodes are implemented as part of the mutation. In addition to crossover and mutation, a bank of genes representing segments of various search movements is used as a part of the process of evolution.

Lagrangian and Eulerian approaches are used to describe the particle motion for the modeling of the moving target distribution. finally, we concentrate on the tactical advantage of using multiple searchers against stationary and evasive targets in a simple environment.

We present a simple example to illustrate, via a simulation, that the developed algorithm, OASPP(Optimal Acoustic Search Path Planning), produces the optimal search path for the case when the intuitive solution exists.