Perception, Guidance, and Navigation for Indoor Autonomous Drone Racing Using Deep Learning

By Sunggoo Jung, Sunyou Hwang, Heemin Shin, David Hyunchul Shim. 2018

170 citations recorded by the source corpus.

In autonomous drone racing, a drone is required to fly through the gates quickly without any collision. Therefore, it is important to detect the gates reliably using computer vision. However, due to the complications such as varying lighting conditions and gates seen overlapped, traditional image processing algorithms based on color and geometry of the gates tend to fail during the actual racing. In this letter, we introduce a convolutional neural network to estimate the center of a gate robustly. Using the detection results, we apply a line-of-sight guidance algorithm. The proposed algorithm is implemented using low cost, off-the-shelf hardware for validation. All vision processing is performed in real time on the onboard NVIDIA Jetson TX2 embedded computer. In a number of tests our proposed framework successfully exhibited fast and reliable detection and navigation performance in indoor environment.

Subfields: Aerospace Engineering, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Aerospace Engineering

In Computer Vision and Pattern Recognition

Citations in the corpus

Beauty and the Beast: Optimal Methods Meet Learning for Drone Racing (2018, methodology): “Recent work has shown that deep networks can provide drones with robust perception capabilities and facilitate safe navigation even in dynamic environments [9], [8].”
Deep Drone Racing: Learning Agile Flight in Dynamic Environments (2018, background): “While being independent of any global map and position estimate, these methods are not directly applicable to our specific problem due to their high computational complexity [27, 11], their low maximum speed [13] or the inherent difficulties of generalizing to 3D motions [25, 4, 24].”
Combining Optimal Control and Learning for Visual Navigation in Novel Environments (2019, background): “lly feasible low-level control for goal-driven behavior. Combining Optimal Control and Learning. A number of papers seek to combine the best of learning with optimal control for high-speed navigation [27, 28, 29, 30, 31]. Drews et al. [28, 32] learn a cost function from monocular images for aggressi”