Developing generative adversarial nets to extend training sets and optimize diiscrete actions

• Autorzy: Zhang R. L. , Furusho M.

Identyfikatory

DOI

10.12716/1001.13.04.22

• Języki publikacji: EN

Abstrakty

EN

This study proposes the use of generative adversarial networks (GANs) to solve two crucial problems in the unmanned ship navigation: insufficient training data for neural networks and convergence of optimal actions under discrete conditions. To achieve smart collision avoidance of unmanned ships in various sea environments, first, this study proposes a collision avoidance decision model based on a deep reinforcement learning method. Then, it utilizes GANs to generate enough realistic image training sets to train the decision model. According to generative network learning, the conditional probability distribution of ship maneuvers is learnt (action units). Subsequently, the decision system can select a reasonable action to avoid the obstacles due to the discrete responses of the generated model to different actions and achieve the effect of intelligent collision avoidance. The experimental results showed that the generated target ship image set can be used as the training set of decision neural networks. Further, a theoretical reference to optimize the optimal convergence of discrete actions is provided.

Słowa kluczowe

EN

Maritime Education and Training (MET); Generative Adversarial Network (GAN); discrete actions; MET System in Japan; Lifeboat; Monte Carlo Tree Search (MCTS); learning methods; unmanned ship navigation

• Wydawca: Faculty of Navigation, Gdynia Maritime University
• Czasopismo: TransNav : International Journal on Marine Navigation and Safety of Sea Transportation
• Rocznik: 2019
• Tom: Vol. 13. no. 4
• Strony: 875--880
• Opis fizyczny: Bibliogr. 10 poz., rys.

Twórcy

autor

Zhang R. L.

• Kobe University, Kobe, Japan

autor

Furusho M.

• Kobe University, Kobe, Japan

Bibliografia

• 1. Capraro,  G.  T.,  Farina,  A.,  Griffiths,  H.  &  Wicks,  M.  C.  (2006).  Knowledge‐based  radar  signal  and  data  processing:  a  tutorial  review.  IEEE  Signal  Processing  Magazine, 23(1), 18‐29. 
• 2. Finn,  A.  &  Scheding,  S.  (2010).  Developments  and  challenges  for  autonomous  unmanned  vehicles.  Intelligent Systems Reference Library, 3, 128‐154.
• 3. Goodfellow, I., Pouget‐Abadie, J., Mirza, M., Xu, B., WardeFarley, D., Ozair, S. ... & Bengio, Y. (2014). Generative  adversarial  nets.  In  Advances  in  neural  information  processing systems (pp. 2672‐2680).
• 4. Mazurowski, M. A., Habas, P. A., Zurada, J. M., Lo, J. Y.,  Baker, J. A. & Tourassi, G. D. (2008). Training neural  network  classifiers  for  medical  decision  making:  The effects  of  imbalanced  datasets  on  classification  performance. Neural networks, 21(2‐3), 427‐436
• 5. Mnih, V., Kavukcuoglu, K., Silver, D., Rusu, A. A., Veness,  J., Bellemare, M. G. ... & Petersen, S. (2015). Human‐level  control  through  deep  reinforcement  learning.  Nature,  518(7540), 529
• 6. Sarukkai, R. R. (2000). Link prediction and path analysis  using  Markov  chains1.  Computer  Networks,  33(1‐6),  377‐386
• 7. Scheffer, M., Carpenter, S. R., Lenton, T. M., Bascompte, J.,  Brock,  W.,  Dakos,  V.,  ...  &  Pascual,  M.  (2012).  Anticipating critical transitions. science, 338(6105), 344348. 
• 8. Trucco, P., Cagno, E., Ruggeri, F. & Grande, O. (2008). A  Bayesian  Belief  Network  modelling  of  organisational  factors  in  risk  analysis:  A  case  study  in  maritime  transportation. Reliability Engineering & System Safety,  93(6), 845‐856
• 9. Wang,  X.,  Yadav,  V.  &  Balakrishnan,  S.  N.  (2007).  Cooperative  UAV  formation  flying  with  obstacle/collision avoidance. 
• 10. Zhang, R. L. & Furusho, M. (2017). Conversion timing of  seafarer’s  decision‐making  for  unmanned  ship  navigation. TransNav: International Journal on Marine  Navigation and Safety of Sea Transportation,