Double-Prong ConvLSTM for Spatiotemporal Occupancy Prediction in Dynamic Environments

TL;DR

This paper introduces a double-prong neural network architecture that improves long-term spatiotemporal occupancy prediction for autonomous vehicles by effectively modeling static and dynamic environment components.

Contribution

The proposed double-prong ConvLSTM architecture uniquely combines static environment and dynamic object predictions, enhancing long-term occupancy forecasting accuracy.

Findings

Fused output retains dynamic objects effectively.

Reduces blurriness in long-horizon predictions.

Outperforms baseline models on Waymo dataset.

Abstract

Predicting the future occupancy state of an environment is important to enable informed decisions for autonomous vehicles. Common challenges in occupancy prediction include vanishing dynamic objects and blurred predictions, especially for long prediction horizons. In this work, we propose a double-prong neural network architecture to predict the spatiotemporal evolution of the occupancy state. One prong is dedicated to predicting how the static environment will be observed by the moving ego vehicle. The other prong predicts how the dynamic objects in the environment will move. Experiments conducted on the real-world Waymo Open Dataset indicate that the fused output of the two prongs is capable of retaining dynamic objects and reducing blurriness in the predictions for longer time horizons than baseline models.