Dynamics-Aware Spatiotemporal Occupancy Prediction in Urban Environments

Abstract

For autonomous vehicles to proactively plan safe trajectories and make informed decisions, they must be able to predict the future occupancy states of the local environment. However, common issues with occupancy prediction include predictions where moving objects vanish or become blurred, particularly at longer time horizons. We propose an environment prediction framework that incorporates environment semantics for future occupancy prediction. Our method first semantically segments the environment and uses this information along with the occupancy information to predict the spatiotemporal evolution of the environment. We validate our approach on the real-world Waymo Open Dataset. Compared to baseline methods, our model has higher prediction accuracy and is capable of maintaining moving object appearances in the predictions for longer prediction time horizons.

UFOMap: An Efficient Probabilistic 3D Mapping Framework That Embraces the Unknown

“…Mersch et al [3], Sun et al [4], and Toyungyernsub et al [5] develop novel frameworks to extract features and detect dynamic points utilizing spatial and temporal information. Some of these methods use the point cloud format, while others choose to translate point clouds into different representations, such as residual images, to facilitate processing.…”

PointMapNet: Point Cloud Feature Map Network for 3D Human Action Recognition

“…Many existing methods utilize the high-level history state information (e.g., position, velocity) and the context information (e.g., roadgraph/map, context agent trajectory) to forecast future state sequences [5], [7], [8], [10], [29]- [36]. There are two widely used ways to represent the roadgraph information: (a) rasterized top-down view images [29], [36], [37]; and (b) roadgraph vectors [5], [32]. In order to model the interactions between entities, different feature aggregation techniques are employed such as social pooling [35], attention mechanisms [7], and message passing across graphs [8].…”

Allo-centric Occupancy Grid Prediction for Urban Traffic Scene Using Video Prediction Networks

“…Recent works have considered the problem of future egocentric OGM predictions ( [6], [7], [8], [9], [10], [3]) by incorporating spatio-temporal deep-learning methods, involving combinations of convolutional neural networks (CNNs) and recurrent neural networks (RNNs). These works predict the complete scene as OGMs and encounter similar challenges for forecasting such as blurriness, loss of scene Fig.…”

“…These grid predictions, however, present challenges of blurry dynamic vehicles and their evaluation. With few common benchmark and direct comparison methods being available, the future prediction of OGMs are mostly evaluated against the actual generated grid ( [3], [4]) rather than the ground truth. This carries the inherent risk of overlooking potential errors in OGMs generation during training and 1 Univ.…”