Space, Time, and Interaction: A Taxonomy of Corner Cases in Trajectory Datasets for Automated Driving

Abstract

Trajectory data analysis is an essential component for highly automated driving. Complex models developed with these data predict other road users' movement and behavior patterns. Based on these predictions -and additional contextual information such as the course of the road, (traffic) rules, and interaction with other road users -the highly automated vehicle (HAV) must be able to reliably and safely perform the task assigned to it, e.g., moving from point A to B. Ideally, the HAV moves safely through its environment, just as we would expect a human driver to do. However, if unusual trajectories occur, so-called trajectory corner cases, a human driver can usually cope well, but an HAV can quickly get into trouble. In the definition of trajectory corner cases, which we provide in this work, we will consider the relevance of unusual trajectories with respect to the task at hand. Based on this, we will also present a taxonomy of different trajectory corner cases. The categorization of corner cases into the taxonomy will be shown with examples and is done by cause and required data sources. To illustrate the complexity between the machine learning (ML) model and the corner case cause, we present a general processing chain underlying the taxonomy.

Corner cases in machine learning processes

“…Relevance represents the importance or significance of something to achieve a specific goal. In terms of ML, it indicates the importance or significance of a data sample, e.g., corner case, for an application task at hand [28]. A cyclist, only a few meters in front of the car, is more relevant for object detection from the safety point of view than a cyclist far away.…”

“…A corner case definition is given in [28,32] for trajectory data. Their definitions are not restricted to a single trajectory alone.…”

“…However, what do we actually mean by a corner case? -We have noticed that everyone has their interpretation and understanding [24][25][26][27][28]. This includes corner case examples, associated properties, and definitions regarding a corner case, some of which can be very different.…”

“…How to define the term corner cases with respect to ML is not yet conclusively clarified. In the literature, there are some starting points from software and hardware testing [24,25] and some definitions for ML perspective [26][27][28][29][30][31][32][33][34][35].…”

“…However, in particular, we also look at each of the listed properties from the perspective of an ML model. Definitions of corner cases are not new but usually have a strong connection to an application [26][27][28]. Instead, we aim for a more general corner case description, which is mainly based on the particularities and properties of a corner case.…”