A Directed Graph and GRUs-Based Trajectory Forecasting of Intelligent and Automated Transportation System for Consumer Electronics

Abstract

Recent technological developments led to a rapid increase in the use of consumer electronic products. Presently, the car has gained more attention as an electronic device due to its sophistication. Trajectory forecasting is crucial for the decision-making system of the intelligent transportation system. The challenging task for intelligent vehicles and even human drivers is to provide the interaction. In this manuscript, the proposed model has been discussed, that extends the spectral graph into the directed graph with first and second-order approximation. This retains the interaction between the nodes of the directed graph and expands the receptive field for the convolution. The extracted features from the directed graph convolutional network is passing through the multiple Gated Recurrent Units (GRUs) with skip connection trajectory prediction of the dynamic agents. The invented model is trained with three publically available datasets: Argoverse, Apolloscape and Lyft, which demonstrate the superiority of the proposed model over the existing state-of-the-art methods.