Driving Policy Transfer via Modularity and Abstraction

Abstract

End-to-end approaches to autonomous driving have high sample complexity and are difficult to scale to realistic urban driving. Simulation can help end-to-end driving systems by providing a cheap, safe, and diverse training environment. Yet training driving policies in simulation brings up the problem of transferring such policies to the real world. We present an approach to transferring driving policies from simulation to reality via modularity and abstraction. Our approach is inspired by classic driving systems and aims to combine the benefits of modular architectures and end-to-end deep learning approaches. The key idea is to encapsulate the driving policy such that it is not directly exposed to raw perceptual input or low-level vehicle dynamics. We evaluate the presented approach in simulated urban environments and in the real world. In particular, we transfer a driving policy trained in simulation to a 1/5-scale robotic truck that is deployed in a variety of conditions, with no finetuning, on two continents. The supplementary video can be viewed at https://youtu.be/BrMDJqI6H5U

arXiv commentAccepted at Conference on Robotic Learning (CoRL'18) http://proceedings.mlr.press/v87/mueller18a.html

Cited by in corpus (39)

• On Evaluation of Embodied Navigation Agents
• Multimodal End-to-End Autonomous Driving
• A Survey of End-to-End Driving: Architectures and Training Methods
• A General Framework for Uncertainty Estimation in Deep Learning
• Deep Drone Racing: From Simulation to Reality with Domain Randomization
• A Survey of Deep RL and IL for Autonomous Driving Policy Learning
• Sim2Real Predictivity: Does Evaluation in Simulation Predict Real-World Performance?
• LookOut: Diverse Multi-Future Prediction and Planning for Self-Driving
• Deep Local Trajectory Replanning and Control for Robot Navigation
• Learned Visual Navigation for Under-Canopy Agricultural Robots
• VTGNet: A Vision-based Trajectory Generation Network for Autonomous Vehicles in Urban Environments
• ReLMoGen: Leveraging Motion Generation in Reinforcement Learning for Mobile Manipulation
• BEV-Seg: Bird's Eye View Semantic Segmentation Using Geometry and Semantic Point Cloud
• Combining Optimal Control and Learning for Visual Navigation in Novel Environments
• Learning to Set Waypoints for Audio-Visual Navigation
• Learning Robotic Navigation from Experience: Principles, Methods, and Recent Results
• Sim2real for Autonomous Vehicle Control using Executable Digital Twin
• Learning Functionally Decomposed Hierarchies for Continuous Control Tasks with Path Planning
• Navigation-Oriented Scene Understanding for Robotic Autonomy: Learning to Segment Driveability in Egocentric Images
• Situational Fusion of Visual Representation for Visual Navigation
• SAM: Squeeze-and-Mimic Networks for Conditional Visual Driving Policy Learning
• Domain Adaptation Through Task Distillation
• Adversarial Evaluation of Autonomous Vehicles in Lane-Change Scenarios
• DSDNet: Deep Structured self-Driving Network
• Simulation to Scaled City: Zero-Shot Policy Transfer for Traffic Control via Autonomous Vehicles
• Motion Perception in Reinforcement Learning with Dynamic Objects
• AdvSim: Generating Safety-Critical Scenarios for Self-Driving Vehicles
• Generating Robust Supervision for Learning-Based Visual Navigation Using Hamilton-Jacobi Reachability
• Deep Object-Centric Policies for Autonomous Driving
• Vision-based DRL Autonomous Driving Agent with Sim2Real Transfer
• Improving Robustness of Deep Neural Networks for Aerial Navigation by Incorporating Input Uncertainty
• Learning 3D-aware Egocentric Spatial-Temporal Interaction via Graph Convolutional Networks
• Pre-training of Deep RL Agents for Improved Learning under Domain Randomization
• Deep Structured Reactive Planning
• Investigation of Factorized Optical Flows as Mid-Level Representations
• GAPLE: Generalizable Approaching Policy LEarning for Robotic Object Searching in Indoor Environment
• Safer End-to-End Autonomous Driving via Conditional Imitation Learning and Command Augmentation
• Expert Human-Level Driving in Gran Turismo Sport Using Deep Reinforcement Learning with Image-based Representation
• Few-Shot Image-to-Semantics Translation for Policy Transfer in Reinforcement Learning