View article

A better understanding of agents’ behaviour in a dynamic traffic environment is required for an efficient modelling and navigation of autonomous vehicles. In this project we plan to address the problem of motion forecasting of traffic actors through experimentation on the Argoverse Motion Forecasting dataset. We attempt to tackle this challenge using generative adversarial networks (GANs) and compare out results with baseline methods of seq-to-seq prediction and social LSTM provided by the Argoverse Challenge.