Modeling Human Car-Following Behavior from Demonstration with Recurrent Neural Networks
|Modeling Human Car-Following Behavior from Demonstration with Recurrent Neural Networks
|Year of Publication
|Jones, I, Walter, M
|car-following models, driving behavior, recurrent neural network, simulation, trajectory prediction
The validity of simulation testing for autonomous vehicles depends on the ability to accurately simulate human driving behavior. This project seeks to train a model on an individual’s driving data, and to test the ability of the model to predict trajectories that replicate the driver’s style by using the model in a realistic simulated environment. Specifically, we deployed Recurrent Neural Network (RNN) modeling techniques to create a black-box model of an individual’s driving behavior. We use our RNN-trained model to simulate a human-driven vehicle in the Robot Operating System (ROS) based CAT Vehicle simulator for autonomous vehicle validation. We hope this work is a step to improve testing environments for validating human behavior replicating car-following models and thereby improve testing environments for autonomous vehicles in general.