Towards Defensive Autonomous Driving: Collecting and Probing Driving Demonstrations of Mixed Qualities

TL;DR

This paper introduces the R3 Driving Dataset with diverse abnormal driving behaviors to improve safety and out-of-distribution detection in autonomous driving systems, demonstrating the effectiveness of uncertainty and anomaly detection methods.

Contribution

The paper presents a novel dataset with varied abnormal driving scenarios and evaluates detection methods, advancing safety in autonomous driving.

Findings

Uncertainty estimation and anomaly detection can effectively identify abnormal driving behaviors.

The dataset includes 8 categories and 369 detailed abnormal situations.

Most abnormal cases can be discriminated using the proposed methods.

Abstract

Designing or learning an autonomous driving policy is undoubtedly a challenging task as the policy has to maintain its safety in all corner cases. In order to secure safety in autonomous driving, the ability to detect hazardous situations, which can be seen as an out-of-distribution (OOD) detection problem, becomes crucial. However, most conventional datasets only provide expert driving demonstrations, although some non-expert or uncommon driving behavior data are needed to implement a safety guaranteed autonomous driving platform. To this end, we present a novel dataset called the R3 Driving Dataset, composed of driving data with different qualities. The dataset categorizes abnormal driving behaviors into eight categories and 369 different detailed situations. The situations include dangerous lane changes and near-collision situations. To further enlighten how these abnormal driving behaviors can be detected, we utilize different uncertainty estimation and anomaly detection methods to the proposed dataset. From the results of the proposed experiment, it can be inferred that by using both uncertainty estimation and anomaly detection, most of the abnormal cases in the proposed dataset can be discriminated. The dataset of this paper can be downloaded from https://rllab-snu.github.io/projects/R3-Driving-Dataset/doc.html.