Implementation of Neural Networks for Car Navigation System Development in a 2D Car Game Simulation Using Genetic Algorithms
Article Sidebar
Main Article Content
Abstract
Cars in a racing game can be driven by a small neural network whose weights are evolved by a genetic algorithm instead of being programmed or trained on labeled data. This study documents such a system built in Unity: each car reads five distance sensors, passes them through a 5–4–3–2 network (47 evolved values), and is evolved with a genetic algorithm that keeps the two best of 20 cars as parents. We re-examine the evaluation reported for it. Ten of eleven functional tests passed; the network and sensor display failed. On three tracks the share of cars finishing rose from 30%, 20%, and 10% in generations 1–10 to 70%, 50%, and 30% in generations 41–50, a growth of 1.00, 0.75, and 0.50 percentage points per generation that is almost perfectly linear (R² ≥ 0.98), and the open-ended window “51+” reaches 100%, 60%, and 40%. A generation took about 15 s and showed no trend over five generations. Mean FPS fell by 0.20 (high-end PC) and 0.32 (low-end PC) per added car, and only the low-end PC at 100 cars reached the 30 FPS limit. Several reported values are mutually inconsistent, including identical memory columns for the two PCs that the accompanying chart contradicts. The system shows that the pipeline works on the tracks used for training; evidence of learning that generalizes, and of its cost, needs repeated runs, raw logs, and a held-out track.
Downloads
Article Details
Section
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.
How to Cite
References
[1] S. Risi and J. Togelius, “Neuroevolution in games: State of the art and open challenges,” IEEE Trans. Comput. Intell. AI Games, vol. 9, no. 1, pp. 25–41, 2017, doi: 10.1109/TCIAIG.2015.2494596.
[2] J. Schmidhuber, “Deep learning in neural networks: An overview,” Neural Netw., vol. 61, pp. 85–117, 2015, doi: 10.1016/j.neunet.2014.09.003.
[3] V. Mnih et al., “Human-level control through deep reinforcement learning,” Nature, vol. 518, no. 7540, pp. 529–533, 2015, doi: 10.1038/nature14236.
[4] D. Whitley, T. Starkweather, and C. Bogart, “Genetic algorithms and neural networks: Optimizing connections and connectivity,” Parallel Comput., vol. 14, no. 3, pp. 347–361, 1990, doi: 10.1016/0167-8191(90)90086-O.
[5] X. Yao, “Evolving artificial neural networks,” Proc. IEEE, vol. 87, no. 9, pp. 1423–1447, 1999, doi: 10.1109/5.784219.
[6] D. Floreano, P. Dürr, and C. Mattiussi, “Neuroevolution: From architectures to learning,” Evol. Intell., vol. 1, no. 1, pp. 47–62, 2008, doi: 10.1007/s12065-007-0002-4.
[7] R. Mahajan and G. Kaur, “Neural networks using genetic algorithms,” Int. J. Comput. Appl., vol. 77, no. 14, pp. 6–11, 2013, doi: 10.5120/13549-1153.
[8] L. R. Manangka, H. Suprijono, and D. Nurcipto, “Pengenalan pola lintasan berbasis neural network pada prototype self-driving car” (in Indonesian), Elektrika, vol. 12, no. 2, p. 67, 2020, doi: 10.26623/elektrika.v12i2.2732.
[9] Y. B. Setiawanda, M. K. Rasyid, M. J. Ramadhan, and A. D. Hartanto, “Penerapan algoritma genetika dan algoritma ghost framework pada game Ms. Pacman” (in Indonesian), Creat. Inf. Technol. J., vol. 5, no. 3, p. 174, 2019, doi: 10.24076/citec.2018v5i3.206.
[10] B. A. Qader, K. H. Jihad, and M. R. Baker, “Evolving and training of neural network to play DAMA board game using NEAT algorithm,” Informatica, vol. 46, no. 5, pp. 29–37, 2022, doi: 10.31449/inf.v46i5.3897.
[11] M. Tjahyadi, A. Sinsuw, V. Tulenan, and S. Sentinuwo, “Prototipe game musik bambu menggunakan engine Unity 3D” (in Indonesian), J. Tek. Inform., vol. 4, no. 2, 2015, doi: 10.35793/jti.4.2.2014.6990.
[12] S. Katoch, S. S. Chauhan, and V. Kumar, “A review on genetic algorithm: Past, present, and future,” Multimed. Tools Appl., vol. 80, no. 5, pp. 8091–8126, 2021, doi: 10.1007/s11042-020-10139-6.
[13] T. Blickle and L. Thiele, “A comparison of selection schemes used in evolutionary algorithms,” Evol. Comput., vol. 4, no. 4, pp. 361–394, 1996, doi: 10.1162/evco.1996.4.4.361.
[14] A. Lipowski and D. Lipowska, “Roulette-wheel selection via stochastic acceptance,” Phys. A Stat. Mech. Appl., vol. 391, no. 6, pp. 2193–2196, 2012, doi: 10.1016/j.physa.2011.12.004.
[15] S. Nidhra, “Black box and white box testing techniques – A literature review,” Int. J. Embedded Syst. Appl., vol. 2, no. 2, pp. 29–50, 2012, doi: 10.5121/ijesa.2012.2204.
[16] L. D. Brown, T. T. Cai, and A. DasGupta, “Interval estimation for a binomial proportion,” Stat. Sci., vol. 16, no. 2, pp. 101–133, 2001, doi: 10.1214/ss/1009213286.