Abstract:In this paper, we train a recurrent neural network to learn dynamics of a chaotic road environment and to project the future of the environment on an image. Future projection can be used to anticipate an unseen environment for example, in autonomous driving. Road environment is highly dynamic and complex due to the interaction among traffic participants such as vehicles and pedestrians. Even in this complex environment, a human driver is efficacious to safely drive on chaotic roads irrespective of the number of traffic participants. The proliferation of deep learning research has shown the efficacy of neural networks in learning this human behavior. In the same direction, we investigate recurrent neural networks to understand the chaotic road environment which is shared by pedestrians, vehicles (cars, trucks, bicycles etc.), and sometimes animals as well. We propose \emph{Foresee}, a unidirectional gated recurrent units (GRUs) network with attention to project future of the environment in the form of images. We have collected several videos on Delhi roads consisting of various traffic participants, background and infrastructure differences (like 3D pedestrian crossing) at various times on various days. We train \emph{Foresee} in an unsupervised way and we use online training to project frames up to $0.5$ seconds in advance. We show that our proposed model performs better than state of the art methods (prednet and Enc. Dec. LSTM) and finally, we show that our trained model generalizes to a public dataset for future projections.

Dynamic Environment Prediction in Urban Scenes using Recurrent Representation Learning

Dynamic Occupancy Grid Prediction for Urban Autonomous Driving: A Deep Learning Approach with Fully Automatic Labeling

StopNet: Scalable Trajectory and Occupancy Prediction for Urban Autonomous Driving

Learning Structured Representations of Spatial and Interactive Dynamics for Trajectory Prediction in Crowded Scenes

Predicting Future Spatiotemporal Occupancy Grids with Semantics for Autonomous Driving

Path Planning in Dynamic Environments through Trajectory Prediction and Reinforcement Learning

LOPR: Latent Occupancy PRediction using Generative Models

RPP-Net: Rigid Constrained Point Cloud Prediction Network

A Multi-Task Recurrent Neural Network for End-to-End Dynamic Occupancy Grid Mapping

Deep Occupancy-Predictive Representations for Autonomous Driving

PredictionNet: Real-Time Joint Probabilistic Traffic Prediction for Planning, Control, and Simulation

Occupancy Map Prediction Using Generative and Fully Convolutional Networks for Vehicle Navigation

Deep Tracking on the Move: Learning to Track the World from a Moving Vehicle using Recurrent Neural Networks

Foresee: Attentive Future Projections of Chaotic Road Environments with Online Training

Self-supervised Multi-future Occupancy Forecasting for Autonomous Driving

Social LSTM: Human Trajectory Prediction in Crowded Spaces

Learning-based 3D Occupancy Prediction for Autonomous Navigation in Occluded Environments

Multiagent Multimodal Trajectory Prediction in Urban Traffic Scenarios Using a Neural Network-Based Solution

Probabilistic vehicle trajectory prediction over occupancy grid map via recurrent neural network

Learning Occupancy Priors of Human Motion from Semantic Maps of Urban Environments