Simultaneous Policy Learning and Latent State Inference for Imitating Driver Behavior

TL;DR

This paper introduces a method for learning driver behavior models that infer unobserved variables and improve imitation by jointly learning policies and latent state encodings, demonstrating effectiveness on synthetic data.

Contribution

The work presents a novel approach for simultaneous policy learning and latent state inference for driver behavior modeling without prior knowledge of driver classes.

Findings

Models learn to distinguish four driver behavior classes.

Policies with latent variables outperform baselines in imitation tasks.

Actions are heavily influenced by inferred latent states.

Abstract

In this work, we propose a method for learning driver models that account for variables that cannot be observed directly. When trained on a synthetic dataset, our models are able to learn encodings for vehicle trajectories that distinguish between four distinct classes of driver behavior. Such encodings are learned without any knowledge of the number of driver classes or any objective that directly requires the models to learn encodings for each class. We show that driving policies trained with knowledge of latent variables are more effective than baseline methods at imitating the driver behavior that they are trained to replicate. Furthermore, we demonstrate that the actions chosen by our policy are heavily influenced by the latent variable settings that are provided to them.

Full text

See pages 1-7 of latent_driver.pdf