Abstract
Recently, several deep-learning based navigation methods have been achieved because of a high quality dataset collected from high-quality simulated environments. However, the cost of creating high-quality simulated environments is high. In this paper, we present a concept of the reduced simulation, which can serve as a simplified version of a simulated environment yet be efficient enough for training deep-learning based UAV collision avoidance approaches. Our approach deals with the reality gap between a reduced simulation dataset and real world dataset and can provide a clear guideline for reduced simulation design. Our experimental result confirmed that the reduction in visual features provided by textures and lighting does not affect operating performance with the user study. Moreover, by conducting collision detection experiments, we verified that our reduced simulation outperforms the conventional cost-effective simulations in adaptation capability with respect to realistic simulation and real-world scenario.
Highlights
One important insight derived from this question: What is the essential or minimal visual information required to realize Micro aerial vehicles (MAVs) visual control? In this paper, we propose a concept of reduced simulation, which is a data generation scheme for training a machine learning model
In this study, based on training in reduced simulation, we showed that the MAV can fly in an unknown environment without collision and capture data for post processing of the 3D environment
We proposed the novel real-to-sim concept, reduced simulation, to realize autonomous flight in a narrow or confined environment
Summary
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations. There have been recent and rapid advances taking place in convolutional neural networks (ConvNets) and reinforcement learning, which have adopted deep learning models, where the quality of datasets determines performance Some of these studies [4,5] use data collected from the real environment. The current studies on real-to-sim target the grasping task of the manipulator [13] These factors motivated us to embed non-learning traditional image processing into a real-to-sim approach for MAV applications. We tackled a collision detection problem in narrow or confined environments with a novel real-to-sim concept. This restricted environment has not been addressed in previous studies. Based on the results of our experiment, we provided guidelines for adopting a costsaving reduced simulation for MAV collision detection in cluttered environments
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
