Biological creations adapt to environmental changes. Similarly, can autonomous AI system adapt to the environmental changes? During natural disasters such as floods or cyclones, an autonomous robot might unexpectedly face new conditions such as occlusions from dust, and hence may need to adapt itself. Is it possible for a drone flying into a disaster zone to autonomously evolve itself without any human guidance. Many times autonomous AI systems may be exposed to new conditions that it hasn’t yet been trained. How to provision full autonomy to such autonomous AI?. This is the challenge this paper answers. Disruptions in internet connectivity during disasters add an additional dimension to this challenge. How does the AI on drone self-adapt during disasters? Is it possible to employ Neural Architecture Search (NAS) for autonomously evolving the drone’s intelligence to the new environment?. With internet outages during disasters, is it possible to evolve the AI by evolving the model locally on the drone?. In short, this paper explores how to design autonomous drones that can triumph over disasters, by autonomous evolving the drone intelligence to the new environment using NAS.

Artificial Intelligence, Autonomous systems, Neural Architecture Search, AutoML, Transfer Learning

[1] Zoph, Barret, Vasudevan, Vijay, Shlens, Jonathon, and Le, Quoc V.
“Learning transferable architectures for scalable image recognition”. CVPR, 2018.
[2] Pham, Hieu, Melody Y. Guan, Barret Zoph, Quoc V. Le, and Jeff Dean. "Efficient neural architecture search via parameter sharing." arXiv, 1802.03268, 2018.
[3] Stanley,Jeff, Joel, Risto, "Designing neural networks through neuro evolution.", Nature Machine Intelligence, no.1, pp.24-35, 2019
[4] He et al. "Amc: Automl for model compression and acceleration on mobile devices.", ECCV, pp. 784-800. 2018.
[5] Tan, Mingxing, Bo Chen, Ruoming Pang, Vijay Vasudevan, and Quoc V. Le. "Mnasnet: Platform-aware neural architecture search for mobile.", arXiv, 1807.11626, 2018.
[6] Liu et al. "Progressive neural architecture search." , ECCV, pp. 19-34. 2018.
[7] Hundt, Andrew, Varun, Hager, "sharpDARTS: Faster and More Accurate Differentiable Architecture Search." arXiv, 1903.09900 , 2019.
[8] Dong, Xuanyi, Yi, "Searching for a robust neural architecture in four gpu hours.", CVPR, vol. 1, 2019.
[9] Miikkulainen et al. "Evolving deep neural networks.", Artificial Intelligence in the Age of Neural Networks and Brain Computing, pp. 293-312, Academic Press, 2019.
[10] Liang, Jason, Elliot, Babak, Dan , Karl , Miikkulainen. "Evolutionary Neural AutoML for Deep Learning.", arXiv, 1902.06827, 2019.
[11] Elsken, Thomas, Metzen, Frank, "Neural Architecture Search: A Survey.", Journal of Machine Learning Research 20, no. 55, pp.1-21, 2019
[12] Parimala, Rajkumar, Ruba, Vijayalakshmi, "Challenges and Opportunities with Big Data", International Journal of Scientific Research in Computer Science and Engineering, Vol.5, Issue.5, pp.16-20, 2017
[13] S. Verma, S. K. Rathi, V. S. Rathore , "Earth Observation Satellites Series and its Potentialities", International Journal of Scientific Research in Computer Science and Engineering, Vol.5, Issue.4, pp.49-55, 2017
[14] Delmerico et al, "Are we ready for autonomous drone racing? the UZHFPV drone racing dataset.", ICRA, 2019.
[15] Tamaazousti et all. "Learning more universal representations for transfer-learning." IEEE transactions on pattern analysis and machine intelligence, 2019