Uncertainty-aware Short-term Motion Prediction of Traffic Actors for Autonomous Driving

Abstract

We address one of the crucial aspects necessary for safe and efficient operations of autonomous vehicles, namely predicting future state of traffic actors in the autonomous vehicle's surroundings. We introduce a deep learning-based approach that takes into account a current world state and produces raster images of each actor's vicinity. The rasters are then used as inputs to deep convolutional models to infer future movement of actors while also accounting for and capturing inherent uncertainty of the prediction task. Extensive experiments on real-world data strongly suggest benefits of the proposed approach. Moreover, following completion of the offline tests the system was successfully tested onboard self-driving vehicles.

arXiv commentAccepted for publication at IEEE Winter Conference on Applications of Computer Vision (WACV) 2020

References in corpus (9)

• Adam: A Method for Stochastic Optimization
• Very Deep Convolutional Networks for Large-Scale Image Recognition
• End to End Learning for Self-Driving Cars
• What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision?
• Mastering Chess and Shogi by Self-Play with a General Reinforcement Learning Algorithm
• Show and Tell: Lessons learned from the 2015 MSCOCO Image Captioning Challenge
• Horovod: fast and easy distributed deep learning in TensorFlow
• End-to-End Tracking and Semantic Segmentation Using Recurrent Neural Networks
• Uncertainty-Aware Driver Trajectory Prediction at Urban Intersections

Cited by in corpus (22)

• Overcoming Limitations of Mixture Density Networks: A Sampling and Fitting Framework for Multimodal Future Prediction
• Spatially-Aware Graph Neural Networks for Relational Behavior Forecasting from Sensor Data
• Bounding Box Regression with Uncertainty for Accurate Object Detection
• Learning Lane Graph Representations for Motion Forecasting
• Deep Imitation Learning for Autonomous Driving in Generic Urban Scenarios with Enhanced Safety
• Model-free Deep Reinforcement Learning for Urban Autonomous Driving
• Multimodal Trajectory Predictions for Autonomous Driving using Deep Convolutional Networks
• What-If Motion Prediction for Autonomous Driving
• Multimodal Motion Prediction with Stacked Transformers
• SCALE-Net: Scalable Vehicle Trajectory Prediction Network under Random Number of Interacting Vehicles via Edge-enhanced Graph Convolutional Neural Network
• TPNet: Trajectory Proposal Network for Motion Prediction
• Affordance-based Reinforcement Learning for Urban Driving
• TPCN: Temporal Point Cloud Networks for Motion Forecasting
• Improving Movement Predictions of Traffic Actors in Bird's-Eye View Models using GANs and Differentiable Trajectory Rasterization
• Deep Kinematic Models for Kinematically Feasible Vehicle Trajectory Predictions
• Trajectory Prediction using Equivariant Continuous Convolution
• Collaborative Motion Prediction via Neural Motion Message Passing
• TrafficSim: Learning to Simulate Realistic Multi-Agent Behaviors
• Vehicle trajectory prediction in top-view image sequences based on deep learning method
• UST: Unifying Spatio-Temporal Context for Trajectory Prediction in Autonomous Driving
• Future Video Synthesis with Object Motion Prediction
• Convolutions for Spatial Interaction Modeling