Deep Occupancy-Predictive Representations for Autonomous Driving
On the map: End-to-end autonomous driving → End-to-end system architecture
Abstract (arXiv)
Manually specifying features that capture the diversity in traffic environments is impractical. Consequently, learning-based agents cannot realize their full potential as neural motion planners for autonomous vehicles. Instead, this work proposes to learn which features are task-relevant. Given its immediate relevance to motion planning, our proposed architecture encodes the probabilistic occupancy map as a proxy for obtaining pre-trained state representations. By leveraging a map-aware graph formulation of the environment, our agent-centric encoder generalizes to arbitrary road networks and traffic situations. We show that our approach significantly improves the downstream performance of a reinforcement learning agent operating in urban traffic environments.
Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Placed in End-to-end system architecture, one of 9 directions in End-to-end autonomous driving. Builds on 43 of its 71 references that are in the map. First posted on arXiv 7 Mar 2023.
Builds on
43 of 71 references.
References · 43
- 2014Adam: A Method for Stochastic Optimizationcited by 36,960
- 2017Proximal Policy Optimization Algorithmscited by 14,441
- 2017Graph Attention Networkscited by 8,854
- 2017Neural Message Passing for Quantum Chemistrycited by 3,460
- 2017Modeling Relational Data with Graph Convolutional Networkscited by 1,684
- 2018Graph Neural Networks: A Review of Methods and Applicationscited by 1,571
- 2018Relational inductive biases, deep learning, and graph networkscited by 1,341
- 2018Recurrent World Models Facilitate Policy Evolutioncited by 761
- 2019Heterogeneous Graph Attention Networkcited by 732
- 2017Deep Reinforcement Learning that Matterscited by 727
- 2020Deep Reinforcement Learning for Autonomous Driving: A Surveycited by 698
- 2020Learning Lane Graph Representations for Motion Forecastingcited by 380
31 more
- 2018Learning to Drive in a Daycited by 266
- 2018Graph networks as learnable physics engines for inference and controlcited by 242
- 2021MP3: A Unified Model to Map, Perceive, Predict and Plancited by 211
- 2020Perceive, Predict, and Plan: Safe Motion Planning Through Interpretable Semantic Representationscited by 150
- 2021GOHOME: Graph-Oriented Heatmap Output for future Motion Estimationcited by 146
- 2018Recent Advances in Autoencoder-Based Representation Learningcited by 130
- 2018State Representation Learning for Control: An Overviewcited by 125
- 2020A Survey of End-to-End Driving: Architectures and Training Methodscited by 123
- 2021HOME: Heatmap Output for future Motion Estimationcited by 122
- 2021LaneRCNN: Distributed Representations for Graph-Centric Motion Forecastingcited by 112
- 2021Pretraining Representations for Data-Efficient Reinforcement Learningcited by 105
- 2020Bootstrap Latent-Predictive Representations for Multitask Reinforcement Learningcited by 101
- 2016Label-Free Supervision of Neural Networks with Physics and Domain Knowledgecited by 74
- 2021LaPred: Lane-Aware Prediction of Multi-Modal Future Trajectories of Dynamic Agentscited by 64
- 2022Occupancy Flow Fields for Motion Forecasting in Autonomous Drivingcited by 62
- 2020Predictive Information Accelerates Learning in RLcited by 56
- 2017Dynamic Occupancy Grid Prediction for Urban Autonomous Driving: A Deep Learning Approach with Fully Automatic Labelingcited by 30
- 2018Long-Term Occupancy Grid Prediction Using Recurrent Neural Networkscited by 29
- 2018Multi-Step Prediction of Occupancy Grid Maps with Recurrent Neural Networkscited by 23
- 2019Dynamic Environment Prediction in Urban Scenes using Recurrent Representation Learningcited by 18
- 2019Dynamic Interaction-Aware Scene Understanding for Reinforcement Learning in Autonomous Drivingcited by 14
- 2020Graph Neural Networks and Reinforcement Learning for Behavior Generation in Semantic Environmentscited by 13
- 2021StarNet: Joint Action-Space Prediction with Star Graphs and Implicit Global Frame Self-Attentioncited by 12
- 2020DiGNet: Learning Scalable Self-Driving Policies for Generic Traffic Scenarios with Graph Neural Networkscited by 9
- 2021Self-Supervised Simultaneous Multi-Step Prediction of Road Dynamics and Cost Mapcited by 4
- 2021Large Scale Autonomous Driving Scenarios Clustering with Self-supervised Feature Extractioncited by 4
- 2021Towards Learning Generalizable Driving Policies from Restricted Latent Representationscited by 4
- 2023Geometric Deep Learning for Autonomous Driving: Unlocking the Power of Graph Neural Networks With CommonRoad-Geometriccited by 4
- 2019Deep Grid Net (DGN): A Deep Learning System for Real-Time Driving Context Understandingcited by 2
- 2020Goal-Directed Occupancy Prediction for Lane-Following Actorscited by 2
- 2019Design Space of Behaviour Planning for Autonomous Drivingcited by 1
Built on by
Citing papers per year: 2025: 1.
1 paper.
Similar, not cited
Close in wording, no citation either way
- 2023Occupancy Prediction-Guided Neural Planner for Autonomous Driving
- 2026Risk-Aware Occupancy for Safety-Oriented End-to-End Autonomous Driving
- 2025A Physics-informed End-to-End Occupancy Framework for Motion Planning of Autonomous Vehicles
- 2024Driving in the Occupancy World: Vision-Centric 4D Occupancy Forecasting and Planning via World Models for Autonomous Driving
- 2021Learning Occupancy Priors of Human Motion from Semantic Maps of Urban Environments
- 2025Learning to Sample: Reinforcement Learning-Guided Sampling for Autonomous Vehicle Motion Planning
- 2021Increasing the Efficiency of Policy Learning for Autonomous Vehicles by Multi-Task Representation Learning
- 2023Graph-based Prediction and Planning Policy Network (GP3Net) for scalable self-driving in dynamic environments using Deep Reinforcement Learning