2022
DOI: 10.48550/arxiv.2205.14230
|View full text |Cite
Preprint
|
Sign up to set email alerts
|

Semi-supervised Semantics-guided Adversarial Training for Trajectory Prediction

Abstract: Predicting the trajectories of surrounding objects is a critical task in self-driving vehicles and many other autonomous systems. Recent works demonstrate that adversarial attacks on trajectory prediction, where small crafted perturbations are introduced to history trajectories, may significantly mislead the prediction of future trajectories and ultimately induce unsafe planning. However, few works have addressed enhancing the robustness of this important safety-critical task. In this paper, we present the fir… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1

Citation Types

0
3
0

Year Published

2023
2023
2024
2024

Publication Types

Select...
4

Relationship

0
4

Authors

Journals

citations
Cited by 4 publications
(3 citation statements)
references
References 20 publications
0
3
0
Order By: Relevance
“…Machine learning methods have natural strengths in interaction modeling and prediction [30], [31], [32]. [33] makes discrete behavior decisions for mandatory lane changing based on Bayes classifier and decision trees.…”
Section: B Interaction In Dense Trafficmentioning
confidence: 99%
“…Machine learning methods have natural strengths in interaction modeling and prediction [30], [31], [32]. [33] makes discrete behavior decisions for mandatory lane changing based on Bayes classifier and decision trees.…”
Section: B Interaction In Dense Trafficmentioning
confidence: 99%
“…In a connectivity-enhanced transportation system, vehicles can communicate with each other and/or surrounding infrastructures via dedicated short range communication (DSRC) [14] or cellular vehicle-to-everything (C-V2X) [9]. These communications share important information of vehicles' current states (e.g., location, speed, acceleration) and future intentions (e.g., planned actions and trajectories) that go well beyond the perception and prediction capabilities of individual vehicles, e.g., sharing information that are out of sight of the ego vehicle or intentions that cannot be accurately predicted [12]. Beyond that, vehicles can negotiate and coordinate in distributed manner [33], or follow instructions from a central unit [40] to further optimize the system.…”
Section: Introductionmentioning
confidence: 99%
“…The work in [5] further demonstrates that state-of-the-art trajectory prediction models can be significantly misled by naturallooking but carefully-crafted past trajectory of a certain surrounding vehicle, and discusses several defense methods such as smoothing and SVM-based detection. [6] shows that adversarial training techniques can mitigate the effect of adversarial trajectories. However, few works focus on advanced online anomaly detection methods for vehicle trajectories.…”
Section: Introductionmentioning
confidence: 99%