Roadmap

Autonomous Driving Formula-one and Iris drone with Behavior Metrics and Deep Learning Studio

Abstract

This project is based on JdeRobot Behavior Metrics [1] and DL-Studio [2] frameworks. It aims towards developing the Formula-1 car line following behavior and extending the codebase to PyTorch in addition to already existing Tensorflow support. With this, the preliminary preparations for introducing the drone framework is complete and helps in succesful integration of Iris drone by JdeRobot Drone [3] assets with the current framework. The drone challenge is to solve the same line following problem, however, with some added degrees of freedom which makes it even more challenging. The project explores various State-of-the-art Deep learning [4, 5] based visual control of drones and finally concludes with developing appropriate metrics to evaluate drone brains and expand the existing ones to adapt to the drone framework.

References

[1] https://github.com/JdeRobot/BehaviorMetrics
[2] https://github.com/JdeRobot/DeepLearningStudio
[3] https://github.com/JdeRobot/drones
[4] Bojarski, Mariusz, et al. “End to end learning for self-driving cars.” arXiv:1604.07316 (2016). https://arxiv.org/abs/1604.07316
[5] Rojas-Perez, L.O., & Martinez-Carranza, J. (2020). DeepPilot: A CNN for Autonomous Drone Racing. Sensors, 20(16), 4524. https://doi.org/10.3390/s20164524