Yapay Arı Koloni Algoritması ile Eğitilmiş Tekrarlayıcı Sinir Ağlarının Robot Navigasyonu İçin Kullanılması

Örneklere bağlı olarak dinamik öğrenme yetenekleri sayesinde, doğrusal ve doğrusal olmayan ilişkileri çözümleyerek başarılısonuçlar üreten yapay sinir ağları birçok alanda karşımıza çıkmaktadır. Yapay sinir ağlarında istenen düzeyde performansınsağlanması birçok parametreye bağlı olmakla birlikte, kullanılan ağ modeli ve bu ağın eğitiminde kullanılan algoritmalar üzerindeyapılan çalışmalar giderek artmaktadır. Bu çalışmada, arıların doğada yiyecek arama davranışlarından esinlenilerek geliştirilen yapayarı koloni (Artificial Bee Colony, ABC) algoritması ile eğitilmiş tekrarlayıcı sinir ağlarının (Recurrent Neural Network, RNN) robotnavigasyonunda kullanımına yönelik yeni bir tasarım önerilmiştir. Robotun kontrol stratejisi için üzerine yerleştirilen 24 adetultrasonik sensörden elde edilen veriler kullanılmıştır. Literatürdeki benzer çalışmalarla karşılaştırmak için ortalama karesel hatanınkarekökü ve simetrik oransal ortalama mutlak hata ölçüm metrikleri kullanılmıştır. Elde edilen sonuçlar, önerilen tasarım modelininrobotun hareket yönünün tayininde etkin bir şekilde kullanılabileceğini göstermiştir. Özellikle çok sayıda sensör kullanıldığındaönerilen modelin performansı diğer modellere nazaran çok daha iyi olmuştur.

Using Recurrent Neural Network Models Trained With Artificial Bee Colony Algorithm for Robot Navigation

Artificial neural networks, which produce successful results by establishing linear and nonlinear relations by their dynamic learning abilities, are used in a wide range of fields. Even if the desired performance of networks depends on too many parameters, the number of researches on networks models and learning algorithms are gradually increasing. In this paper a new recurrent neural network (RNN) trained with artificial bee colony (ABC) algorithm was proposed for robot navigation problem. The RNN network trained with sample dataset, obtained from 24 sensors, was used for control strategy in robot movements. Root mean square error (RMS) and symmetric mean absolute percentage error (SMAPE) evalutian metrics were used to compare the proposed method against state of the art algorithms. The results showed that, the performance of the proposed method in determination of robots motion is good and especially, when a large number of sensors used, the proposed model as better performance than the other models.

PDF

___

[1] Kruse, T., Pandey, A.K., Alami, R.ve Kirsch, A. (2013). Human-aware robot navigation: A survey, Robotics and Autonomous Systems 61(12), 1726–1743.
[2] Katsev, M., Yershova, A., Tovar, B., Ghrist, R. Ve La Valle, S.M. (2011). Mapping and pursuit-evasion strategies for a simple wall-following robot, IEEE Transactions on Robotics 27(1), 113–128.
[3] Hoy, M., Matveev, A.S. ve Savkin, A.V. (2015). Algorithms for collision free navigation of mobile robots in complex cluttered environments: a survey. Robotica, 33(3), 463-497.
[4] Yang L, Qi J, Song D, Xiao J, Han J ve Xia Y. (2016). Survey of robot 3D path planning algorithms. Journal of Control Science and Engineering, 5,76-82.
[5] Patle, B.K., Ganesh B.L., A. Pandey, D.R.K. Parhi, A., Jagadeesh. (2019). A review: On path planning strategies for navigation of mobile robot. Defence Technology, InPress.
[6] Trautman, P, Ma, J., Murray, R.M.ve Krause, A. (2015). Robot navigation in dense human crowds: Statistical models and experimental studies of human–robot cooperation, The International Journal of Robotics Research 34(3), 335–356.
[7] Li, T., Sun, Z., Xu Y. Ve Zhang, B. (2015). Robot navigation based on visual feature perception and Monte Carlo sampling, in: Control and Decision Conference (CCDC), 27th Chinese, IEEE, 3237–3242.
[8] Dash, T., Nayak, T. ve Swain, R.R. (2015). Controlling Wall Following Robot Navigation Based on Gravitational Search and Feed Forward Neural Network, Proceedings of the 2nd International Conference on Perception and Machine Intelligence, 196- 200.
[9] Dash, T., Swain, R.R. ve Nayak, T. (2017). Automaticnavigation of wall-following mobile robot using a hybrid metaheuristic assisted neural network, Data Science, 1-17.
[10]Karaboğa D. ve Akay, B. (2009). A comparative study of Artificial Bee Colony algorithm, Applied Mathematics and Computation, 214(1), 108-132.
[11]Karaboğa, D. ve Akay, B. (2007). Artificial Bee Colony (ABC) Algorithm on Training Artificial Neural Networks. Eskişehir, IEEE 15th Signal Processing and Communications Applications.
[12]Karaboğa D., Akay B. ve Öztürk C. (2008). Artificial bee colony (ABC) optimization algorithm for training feed-forward neural networks, 4th International Conference on Modeling Decisions for Artificial Intelligence, Kitakyushu, Japonya, (4617), 318-321.
[13]Karaboga, D. (2005). An Idea Based on Honey Bee Swarm for Numerical Optimization, Technical Report. Erciyes University. Kayseri.
[14]Catalbas, B. (2015). Recurrent Neural Network Learning With An Application To The Control Of Legged, Bilkent University.
[15]Shimodaira, H. (2002). Time-Series Prediction. Cornelius T. Leondes. Expert Systems The Technology of Knowledge Management and Decision Making for the 21st Century, Academic Press, 4, 1295-1311.
[16]Zhang, Y. et al. (2017). A recurrent neural network approach for visual servoing of manipulators. Macau, China, IEEE International Conference on Information andAutomation (ICIA), 614-619.
[17]Dash, T., Ranjan, S. ve Mishra, G. (2014). Neural network approach to control Wall following robot navigation,In Advanced Communication Control and Computing Technologies (ICACCCT), International Conference, India, 1072-1076