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