ChauffeurNet: Learning to Drive by Imitating the Best and Synthesizing the Worst

Abstract

Our goal is to train a policy for autonomous driving via imitation learning that is robust enough to drive a real vehicle. We find that standard behavior cloning is insufficient for handling complex driving scenarios, even when we leverage a perception system for preprocessing the input and a controller for executing the output on the car: 30 million examples are still not enough. We propose exposing the learner to synthesized data in the form of perturbations to the expert's driving, which creates interesting situations such as collisions and/or going off the road. Rather than purely imitating all data, we augment the imitation loss with additional losses that penalize undesirable events and encourage progress -- the perturbations then provide an important signal for these losses and lead to robustness of the learned model. We show that the ChauffeurNet model can handle complex situations in simulation, and present ablation experiments that emphasize the importance of each of our proposed changes and show that the model is responding to the appropriate causal factors. Finally, we demonstrate the model driving a car in the real world.

arXiv commentVideo results: https://sites.google.com/view/waymo-learn-to-drive

Cited by in corpus (43)

• Deep Multi-modal Object Detection and Semantic Segmentation for Autonomous Driving: Datasets, Methods, and Challenges
• One Thousand and One Hours: Self-driving Motion Prediction Dataset
• Causal Confusion in Imitation Learning
• MultiPath: Multiple Probabilistic Anchor Trajectory Hypotheses for Behavior Prediction
• ViNG: Learning Open-World Navigation with Visual Goals
• Injecting Knowledge in Data-driven Vehicle Trajectory Predictors
• Center-based 3D Object Detection and Tracking
• Ensemble Quantile Networks: Uncertainty-Aware Reinforcement Learning with Applications in Autonomous Driving
• Driving Behavior Explanation with Multi-level Fusion
• 3D Point Cloud Processing and Learning for Autonomous Driving
• A Survey on Autonomous Vehicle Control in the Era of Mixed-Autonomy: From Physics-Based to AI-Guided Driving Policy Learning
• Scenario-Transferable Semantic Graph Reasoning for Interaction-Aware Probabilistic Prediction
• NuPlan: A closed-loop ML-based planning benchmark for autonomous vehicles
• RoboNet: Large-Scale Multi-Robot Learning
• Domain Adaptation Through Task Distillation
• SAM: Squeeze-and-Mimic Networks for Conditional Visual Driving Policy Learning
• Decoder Fusion RNN: Context and Interaction Aware Decoders for Trajectory Prediction
• Efficient Deep Reinforcement Learning with Imitative Expert Priors for Autonomous Driving
• Scene Transformer: A unified architecture for predicting multiple agent trajectories
• Design Space of Behaviour Planning for Autonomous Driving
• DeepRacing: Parameterized Trajectories for Autonomous Racing
• Emergent Road Rules In Multi-Agent Driving Environments
• Motion Prediction using Trajectory Sets and Self-Driving Domain Knowledge
• SceneGen: Learning to Generate Realistic Traffic Scenes
• Safe Trajectory Planning Using Reinforcement Learning for Self Driving
• TrafficSim: Learning to Simulate Realistic Multi-Agent Behaviors
• Learning by Cheating
• SurfelGAN: Synthesizing Realistic Sensor Data for Autonomous Driving
• Deep Latent Competition: Learning to Race Using Visual Control Policies in Latent Space
• Efficient Exploration in Constrained Environments with Goal-Oriented Reference Path
• Fast Risk Assessment for Autonomous Vehicles Using Learned Models of Agent Futures
• Multi-Agent Tensor Fusion for Contextual Trajectory Prediction
• Iterative Imitation Policy Improvement for Interactive Autonomous Driving
• Deep Structured Reactive Planning
• Attentional Bottleneck: Towards an Interpretable Deep Driving Network
• Perception-Prediction-Reaction Agents for Deep Reinforcement Learning
• Driving among Flatmobiles: Bird-Eye-View occupancy grids from a monocular camera for holistic trajectory planning
• Autonomy 2.0: Why is self-driving always 5 years away?
• Learning a Model for Inferring a Spatial Road Lane Network Graph using Self-Supervision
• DriverGym: Democratising Reinforcement Learning for Autonomous Driving
• Learning robust driving policies without online exploration
• Polynomial Trajectory Predictions for Improved Learning Performance
• Automatic Curricula via Expert Demonstrations