Learning to Fly by Crashing

2017/04/19 by Dhiraj Gandhi, Lerrel Pinto, Gandhi, Dhiraj +3 · 2 voices · 23 citations
Computer Science · Engineering · #Artificial intelligence #Computer science #Computer vision #Crash #Drone #Machine learning #Overfitting #Robot Manipulation and Learning #Robotic Path Planning Algorithms #Robotics and Sensor-Based Localization #Simple (philosophy) #cs.CV #cs.LG #cs.RO

paper · pdf · doi:10.48550/arxiv.1704.05588

published in arXiv (Cornell University) (Cornell University)

openalex publication_date 2017/04/19 · arxiv created 2017/04/27 · arxiv updated 2017/04/28 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/05

Abstract

How do you learn to navigate an Unmanned Aerial Vehicle (UAV) and avoid obstacles? One approach is to use a small dataset collected by human experts: however, high capacity learning algorithms tend to overfit when trained with little data. An alternative is to use simulation. But the gap between simulation and real world remains large especially for perception problems. The reason most research avoids using large-scale real data is the fear of crashes! In this paper, we propose to bite the bullet and collect a dataset of crashes itself! We build a drone whose sole purpose is to crash into objects: it samples naive trajectories and crashes into random objects. We crash our drone 11,500 times to create one of the biggest UAV crash dataset. This dataset captures the different ways in which a UAV can crash. We use all this negative flying data in conjunction with positive data sampled from the same trajectories to learn a simple yet powerful policy for UAV navigation. We show that this simple self-supervised model is quite effective in navigating the UAV even in extremely cluttered environments with dynamic obstacles including humans. For supplementary video see: https://youtu.be/u151hJaGKUo

Cited by

• CoDrone: Autonomous Drone Navigation Assisted by Edge and Cloud Foundation Models
• Deep Learning-based Robust Autonomous Navigation of Aerial Robots in Dense Forests
• Morphology-Aware Graph Reinforcement Learning for Tensegrity Robot Locomotion
• Oracle-Guided Masked Contrastive Reinforcement Learning for Visuomotor Policies
• Autonomous UAV Flight Navigation in Confined Spaces: A Reinforcement Learning Approach
• Learning with Training Wheels: Speeding up Training with a Simple Controller for Deep Reinforcement Learning
• Vision-based Navigation of Unmanned Aerial Vehicles in Orchards: An Imitation Learning Approach
• Learning to Navigate Autonomously in Outdoor Environments : MAVNet
• Sim2Real View Invariant Visual Servoing by Recurrent Control

Discussions

• Learn to Fly by Crashing [hn, 10 points, 1 comments]
• Learning to Fly by Crashing [hn, 1 points, 0 comments]

Related

• Generalization through Simulation: Integrating Simulated and Real Data into Deep Reinforcement Learning for Vision-Based Autonomous Flight
• MAVRL: Learn to Fly in Cluttered Environments with Varying Speed
• Optimizing UAV Performance In Adverse Flying Conditions Using Reinforcement Learning
• Learning by Cheating : An End-to-End Zero Shot Framework for Autonomous Drone Navigation
• Teaching UAVs to Race: End-to-End Regression of Agile Controls in Simulation