Spatiotemporal Capsule Neural Network for Vehicle Trajectory Prediction

Through advancement of the Vehicle-to-Everything (V2X) network, road safety, energy consumption, and traffic efficiency can be significantly improved. An accurate vehicle trajectory prediction benefits communication traffic management and network resource allocation for the real-time application of the V2X network. Recurrent neural networks and their variants have been reported in recent research to predict vehicle mobility. However, the spatial attribute of vehicle movement behavior has been overlooked, resulting in incomplete information utilization. To bridge this gap, we put forward for the first time a hierarchical trajectory prediction structure using the capsule neural network (CapsNet) with three sequential components. First, the geographic information is transformed into a grid map presentation, describing vehicle mobility distribution spatially and temporally. Second, CapsNet serves as the core model to embed local temporal and global spatial correlation through hierarchical capsules. Finally, extensive experiments conducted on actual taxi mobility data collected in Porto city (Portugal) and Singapore show that the proposed method outperforms the state-of-the-art methods.

I. Introduction

The rapid advancements in self-driving vehicles, communication transceivers, and smart sensors speed up the progress of the Vehicle-to-Everything (V2X) network, which has the potential to bring forth a safe, intelligent, and efficient transportation system. The success of V2X highly depends on the quality of the communication network. Therefore, an accurate prediction of the vehicle quantity plays an important role in optimizing the communication network [1], [2], [3], [4]. Imagine that a large number of intelligent vehicles swarmed into business areas during rush hours will overwhelm communication and computation resources in a short time, causing safety incidences for autonomous vehicles when the service quality is undermined. The real-time prediction of vehicle traffic in an area restricted by limited communication and computation resources allows for alleviating the challenges raised by time-varying network demands. Further, real-time vehicle trajectory prediction saves time for network resource allocators to timely respond, inferring the possible network traffic consumption and conducting predictive tasks offloading in advance. Therefore, vehicle trajectory prediction critically supports to ensure a more reliable vehicular network and its associated services, such as vehicle security [3], [5], self-driving [6], offloading task decisions [7], [8].