Generalization through Simulation: Integrating Simulated and Real Data into Deep Reinforcement Learning for Vision-Based Autonomous Flight

Abstract

Deep reinforcement learning provides a promising approach for vision-based control of real-world robots. However, the generalization of such models depends critically on the quantity and variety of data available for training. This data can be difficult to obtain for some types of robotic systems, such as fragile, small-scale quadrotors. Simulated rendering and physics can provide for much larger datasets, but such data is inherently of lower quality: many of the phenomena that make the real-world autonomous flight problem challenging, such as complex physics and air currents, are modeled poorly or not at all, and the systematic differences between simulation and the real world are typically impossible to eliminate. In this work, we investigate how data from both simulation and the real world can be combined in a hybrid deep reinforcement learning algorithm. Our method uses real-world data to learn about the dynamics of the system, and simulated data to learn a generalizable perception system that can enable the robot to avoid collisions using only a monocular camera. We demonstrate our approach on a real-world nano aerial vehicle collision avoidance task, showing that with only an hour of real-world data, the quadrotor can avoid collisions in new environments with various lighting conditions and geometry. Code, instructions for building the aerial vehicles, and videos of the experiments can be found at github.com/gkahn13/GtS

First three authors contributed equally. Accepted to ICRA 2019

References in corpus (12)

• Auto-Encoding Variational Bayes
• MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications
• DeCAF: A Deep Convolutional Activation Feature for Generic Visual Recognition
• Deep Reinforcement Learning: An Overview
• Action-Conditional Video Prediction using Deep Networks in Atari Games
• A 64mW DNN-based Visual Navigation Engine for Autonomous Nano-Drones
• EPOpt: Learning Robust Neural Network Policies Using Model Ensembles
• Modular Deep Q Networks for Sim-to-real Transfer of Visuo-motor Policies
• Deep Drone Racing: Learning Agile Flight in Dynamic Environments
• Learning Transferable Policies for Monocular Reactive MAV Control
• Deep Predictive Policy Training using Reinforcement Learning
• VR-Goggles for Robots: Real-to-sim Domain Adaptation for Visual Control

Cited by in corpus (4)

• Mid-Level Visual Representations Improve Generalization and Sample Efficiency for Learning Visuomotor Policies
• Combining Optimal Control and Learning for Visual Navigation in Novel Environments
• Situational Fusion of Visual Representation for Visual Navigation
• Generating Robust Supervision for Learning-Based Visual Navigation Using Hamilton-Jacobi Reachability