Elsevier

Applied Soft Computing

Volume 108, September 2021, 107444
Applied Soft Computing

A hybrid binary grey wolf optimizer for selection and reduction of reference points with extreme learning machine approach on local GNSS/leveling geoid determination

https://doi.org/10.1016/j.asoc.2021.107444Get rights and content

Highlights

  • A new problem for reducing the GNSS/leveling geoid points is introduced.

  • Eight different hybrid approaches are used in the solution of the related problem.

  • 102 measured reference points in Burdur, Turkey are used.

  • By the GWO+ELM approach, 82 reference points are reduced to 8 reference points with almost the same accuracy.

  • The cost and time taken to measure the reference points may reduce with the proposed method.

Abstract

Modeling and optimization from natural phenomena and observations of the physical earth is an extremely important issue. In the light of the developments in computer and artificial intelligence technologies, the applications of learning-based modeling and optimization techniques in all kinds of study fields are increasing. In this research, the applicability of four different state-of-the-art metaheuristic algorithms which are Particle swarm optimization (PSO), Tree-Seed Algorithm (TSA), Artificial Bee Colony (ABC) algorithm, and Grey Wolf Optimizer (GWO), in local GNSS/leveling geoid studies have been examined. The most suitable geoid model has been tried to be obtained by using different reference points via the well-known machine learning algorithms, Artificial Neural Network (ANN) and Extreme Learning Machine (ELM), at the existing GNSS/leveling points in Burdur city of Turkey. In this study, eight different hybrid approaches are proposed by using each metaheuristic algorithm together with machine learning methods. By using these hybrid approaches, the model closest to the minimum number of reference points has been tried to be obtained. Furthermore, the performance of the hybrid approaches has been compared. According to the comparisons, the hybrid approach performed with GWO and ELM has achieved better results than other proposed hybrid approaches. As a result of the research, it has been seen that the most suitable local GNSS/Leveling geoid can be determined with a lower number of reference points in an appropriate distribution.

Introduction

Advances in satellite technologies have enabled us to reach a large amount of data about the earth we live on. It is possible to obtain high accuracy positions from satellites, especially for positioning purposes. While heights obtained by GNSS techniques are geometrically defined ellipsoid heights, orthometric or normal heights that are better suited to the physical earth are used in engineering studies [1], [2], [3], [4], [5]. For this reason, geoid determination in which conversion from ellipsoid height to orthometric height is applied is an important issue [6], [7], [8], [9].

Today, many different methods are used in geoid determination studies, depending on the size of the study area and available data. In the literature, there are many different geometric and gravimetric geoid determination methods, mainly astro-geodetic, gravimetric, and GNSS/leveling [6], [8], [10], [11], [12], [13], [14]. One of these, GNSS/leveling geoid has very important for more accurate geoid in engineering applications instead of national or global geoid with high accuracy. For a high-accuracy GNSS/leveling geoid, reference points that reflect the topographic characteristics of the land, distributed appropriately to the project area, are needed. It is clear that the more geoid reference points, the better the modeling can be. However, too many reference points extend the duration of land surveys, which is a factor affecting the cost of projects. One of the main goals is to determine the minimum number of GNSS/leveling geoids with the same accuracy with optimization in the selection of model points. In this study, the optimization issue in local GNSS/leveling geoid determination with ELM is examined. For this purpose, the effect of different algorithms, different reference numbers, and distribution on local GNSS/leveling geoid modeling with ELM was investigated in a sample study area.

Doganalp and Selvi [15] investigated interpolation methods for local geoid determination. In this study, the accuracies of local geoids modeled with Polynomial interpolation, Least-squares collocation method (LSC), Multiquadric (MQ) interpolation method, and Thin Plate Spline (TPS) method for strip area projects are examined.

Erol B and Erol S [16] investigated Learning-based computing techniques in determining local geoids. Multivariable polynomial regression using least squares adjustment (MPRE), ANN, Adaptive network-based fuzzy inference system (ANFIS), and Wavelet neural networks (WNN) methods were used for modeling by using Istanbul metropolitan triangulation network data. The accuracy of each model was calculated by related test points.

Cakır and Yılmaz [1] studied the usability of polynomials, radial basis functions, and multilayer perceptron neural network methods in determining local geoids. Geoid height values of the test points were calculated by using the local geoids obtained from these methods in the test network of the Kayseri region, and the differences between them were interpreted.

Rabah and Kaloop [17] determined the geoid for Egypt by using the minimum curvature surface technique, and the obtained results were compared with the EGM2008 model.

Karaarslan et al. [18] implemented a local geoid determination study for the Trabzon region. The results obtained from Weighted average interpolation, Polynomial surface regression, and Multiquadratic interpolation methods were compared.

Mitas and Mitasova [19] presented the basic principles and mathematical relations of spatial interpolation methods, and 2-dimensional, 3-dimensional, and 4-dimensional interpolations of elevation were explained with examples in GIS studies.

Optimization is the process of optimizing the objective function under current constraints [20]. The Particle swarm optimization (PSO) algorithm is a population-based heuristic optimization algorithm proposed by Eberhart and Kennedy in 1995, inspired by the pattern of social behaviors of bird and fish swarm in finding food, interactions, and achieving goals [21]. Although the PSO algorithm has been proposed to solve continuous problems [22], [23], it has been modified and used to solve binary [24], [25], [26] and discrete optimization problems [27], [28].

Tree-Seed Algorithm (TSA) is a proposed population-based meta-heuristic algorithm, inspired by the relationship between tree and seed [29]. TSA has been used in optimization problems such as constrained optimization [30], [31], optimal power flow [32], traveling salesman problem [33], training ANN [34], and so forth.

The Artificial Bee Colony (ABC) algorithm inspired the foraging behavior of real honeybees and was proposed by Karaboga in 2005 [35]. The ABC algorithm focuses on the social behavior of three different bees in the colony (working bees, stalking bees, and scout bees) to discover nectar and collect it in the hive. This swarm-based algorithm has been applied in many areas [36], [37], [38], [39], [40].

The hunting approaches as a group and the social hierarchy in the pack are so important for the grey​ wolves to live in nature together. Mirjalili et al. was inspired by the social behaviors of these grey​ wolves in the pack for proposing the GWO algorithm [41]. GWO is a state-of-the-art meta-heuristic algorithm that has been used widely so as to solve optimization problems such as optimal power flow [42], [43], wind speed forecasting [44], pattern recognition [45], feature selection [46], unit commitment [47], parameters estimation [48], [49] in recent years and continues to be developed by many researchers. Binary [50], discrete [51], and multi-objective [52] versions of the GWO have been proposed in the literature.

Due to the slow operation of ANN and updating parameters of the network iteratively during the training phase, single-hidden layer feedforward neural networks, called ELM, that select the input weights and calculates the output weights analytically is proposed by Huang et al. [53]. ELM has been used commonly as a regressor such as solar radiation estimation [54], economic growth estimation [55], sensor-less wind speed prediction [56], and so forth. An also, many metaheuristic and ELM hybrids have been proposed in the literature [57], [58], [59], [60], [61], [62], [63].

As can be seen, the effects of modeling and interpolation methods on local geoid determination results were examined in the studies in the literature. What makes this study different from the studies in the literature is the investigation of the effect of reference point selection on the modeling results in the local geoid modeling process. Because the selection of reference points directly affects the model accuracy. Different models can be obtained especially according to the topography of the study area and the distribution of the points. The selection of model reference points is of vital importance in order to produce the closest model to reality. Besides, performing these operations with a small number of points saves time and cost.

In the literature, it is seen that ELM is frequently used as regressors, especially in prediction problems. ELM has been used as a hybrid with meta-heuristic approaches for different purposes in many studies. In this study, ELM and GWO (GWO + ELM) were used as a hybrid and a more accurate modeling possibility with fewer model points with this proposed approach was investigated.

The remainder of the paper is organized as follows. Section 2 provides material and methods of the related study. The experimental setup and experimental results were presented in Section 3. Finally, the work is concluded in Section 4.

Section snippets

Materials and methods

In this study, in which PSO was investigated in determining local GNSS/leveling geoid with ANN, the existing geodetic network in Burdur province was used (Fig. 1). The data used in the study are latitude, longitude, ellipsoid height (φ,λ,h) obtained from GNSS measurements at 102 points and orthometric height (H) values obtained from nivelman (leveling) measurements. GNSS coordinates were obtained from ITRF96 datum observation at 2005.0 epoch, whereas orthometric heights are determined according

Experimental results and discussion

102 reference points used in this study have different orthometric heights as can be seen from the views presented in Figs. 1, 2, and 3. Of the 102 reference points with different orthometric heights, 82 were separated as training data, while 20 were separated as test data. While determining the test points, 5 groups were created by giving the equal number of element properties and rounding ratio 1 in Map Info Professional 17.0.3 program according to orthometric heights. The expert determined 4

Conclusions

Modeling is an important issue in local GNSS/Leveling geoid studies as in all scientific studies. In this study, the effect of optimization with metaheuristic algorithms on the results in local geoid determination with well-known artificial intelligence techniques was investigated. It was aimed to find fewer reference points to represent the entire area without reducing the accuracy of the model. Of the 102 data, 82 were selected as training data and 20 as test data. In this study, to analyze

CRediT authorship contribution statement

Kemal Tütüncü: Supervision, Conceptualization, Writing - original draft. Mehmet Akif Şahman: Data curation, Software, Visualization. Ekrem Tuşat: Writing - original draft, Resources, Investigation.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

We would like to thank Musevitoglu Surveying Co. for the data and technical support used in this study. The authors wish to thank the Scientific Research Projects Coordinatorship at Selçuk University, and The Scientific and Technological Research Council of Turkey for their institutional supports. This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

References (68)

  • CinarA.C. et al.

    A discrete tree-seed algorithm for solving symmetric traveling salesman problem

    Eng. Sci. Technol.

    (2020)
  • XuX. et al.

    Multi-objective artificial bee colony algorithm for multi-stage resource leveling problem in sharing logistics network

    Comput. Ind. Eng.

    (2020)
  • AslanS. et al.

    A genetic artificial bee colony algorithm for signal reconstruction based big data optimization

    Appl. Soft Comput.

    (2020)
  • XuF. et al.

    Training feed-forward artificial neural networks with a modified artificial bee colony algorithm

    Neurocomputing

    (2020)
  • MirjaliliS. et al.

    Grey wolf optimizer

    Adv. Eng. Softw.

    (2014)
  • SulaimanM.H. et al.

    Using the grey wolf optimizer for solving optimal reactive power dispatch problem

    Appl. Soft Comput.

    (2015)
  • SongJ. et al.

    A novel combined model based on advanced optimization algorithm for short-term wind speed forecasting

    Appl. Energy

    (2018)
  • TuQ. et al.

    Multi-strategy ensemble grey wolf optimizer and its application to feature selection

    Appl. Soft Comput.

    (2019)
  • SongX. et al.

    Grey wolf optimizer for parameter estimation in surface waves

    Soil Dyn. Earthq. Eng.

    (2015)
  • EmaryE. et al.

    Binary grey wolf optimization approaches for feature selection

    Neurocomputing

    (2016)
  • LuC. et al.

    An effective multi-objective discrete grey wolf optimizer for a real-world scheduling problem in welding production

    Adv. Eng. Softw.

    (2016)
  • MirjaliliS. et al.

    Multi-objective grey wolf optimizer: A novel algorithm for multi-criterion optimization

    Expert Syst. Appl.

    (2016)
  • HuangG.-B. et al.

    Extreme learning machine: Theory and applications

    Neurocomputing

    (2006)
  • MilačićL. et al.

    Application of artificial neural network with extreme learning machine for economic growth estimation

    Physica A

    (2017)
  • NikolićV. et al.

    Extreme learning machine approach for sensorless wind speed estimation

    Mechatronics

    (2016)
  • UtomoO.K. et al.

    Automatic sleep stage classification using weighted ELM and PSO on imbalanced data from single lead ECG

    Procedia Comput. Sci.

    (2019)
  • ZhuB. et al.

    Hybrid particle swarm optimization with extreme learning machine for daily reference evapotranspiration prediction from limited climatic data

    Comput. Electron. Agric.

    (2020)
  • ChenS. et al.

    Application of PSO-ELM in electronic system fault diagnosis

  • AhilaR. et al.

    An integrated PSO for parameter determination and feature selection of ELM and its application in classification of power system disturbances

    Appl. Soft Comput.

    (2015)
  • LiG. et al.

    Tuning extreme learning machine by an improved artificial bee colony to model and optimize the boiler efficiency

    Knowl.-Based Syst.

    (2014)
  • KiranM.S.

    The continuous artificial bee colony algorithm for binary optimization

    Appl. Soft Comput.

    (2015)
  • YounisG.

    Local earth gravity/potential modeling using ASCH

    Arab. J. Geosci.

    (2015)
  • LiY.-S. et al.

    Research into GNSS levelling using network RTK in Taiwan

    Surv. Rev.

    (2019)
  • TusatE. et al.

    An investigation of the criteria used to select the polynomial models employed in local GNSS/leveling geoid determination studies

    Arab. J. Geosci.

    (2018)
  • Cited by (0)

    View full text