Model-Predictive Policy Learning with Uncertainty Regularization for Driving in Dense Traffic

TL;DR

This paper introduces a method for learning driving policies from observational data by unrolling a learned environment model and penalizing uncertainty to improve policy robustness in dense traffic scenarios.

Contribution

It proposes a novel uncertainty regularization technique for model-predictive policy learning that enables effective policy training solely from observational data without environment interaction.

Findings

Effective driving policies learned from observational data

Uncertainty regularization improves policy robustness

Method outperforms baseline approaches in dense traffic scenarios

Abstract

Learning a policy using only observational data is challenging because the distribution of states it induces at execution time may differ from the distribution observed during training. We propose to train a policy by unrolling a learned model of the environment dynamics over multiple time steps while explicitly penalizing two costs: the original cost the policy seeks to optimize, and an uncertainty cost which represents its divergence from the states it is trained on. We measure this second cost by using the uncertainty of the dynamics model about its own predictions, using recent ideas from uncertainty estimation for deep networks. We evaluate our approach using a large-scale observational dataset of driving behavior recorded from traffic cameras, and show that we are able to learn effective driving policies from purely observational data, with no environment interaction.