Uncertainty estimation for Cross-dataset performance in Trajectory prediction
This addresses the problem of limited generalizability in trajectory prediction for autonomous driving, though it is incremental as it builds on existing methods.
The paper analyzes the cross-dataset generalizability of two state-of-the-art trajectory prediction methods across four datasets (Argoverse, NuScenes, Interaction, Shifts) to assess transferability and dataset representativeness, and introduces a novel uncertainty estimation method to improve cross-dataset performance.
While a lot of work has been carried on developing trajectory prediction methods, and various datasets have been proposed for benchmarking this task, little study has been done so far on the generalizability and the transferability of these methods across dataset. In this paper, we observe the performance of two of the latest state-of-the-art trajectory prediction methods across four different datasets (Argoverse, NuScenes, Interaction, Shifts). This analysis allows to gain some insights on the generalizability proprieties of most recent trajectory prediction models and to analyze which dataset is more representative of real driving scenes and therefore enables better transferability. Furthermore we present a novel method to estimate prediction uncertainty and show how it could be used to achieve better performance across datasets.