Curb Your Attention: Causal Attention Gating for Robust Trajectory Prediction in Autonomous Driving

2024-09-23

arXiv.org

AI Summary

Autonomous vehicle trajectory prediction is vulnerable to non-causal agent interference, compromising robustness and cross-domain generalization. To address this, we propose the first robust prediction framework integrating causal discovery with gated attention: (1) a Causal Discovery Network (CDN) automatically infers the inter-agent causal structure; (2) a Causal Attention Gating (CAG) mechanism is embedded within the Transformer architecture to dynamically suppress propagation of non-causal information. The method operates end-to-end without requiring hand-crafted causal priors, enabling fully data-driven causal modeling. Evaluated on nuScenes and Argoverse 2 benchmarks, our approach improves prediction robustness by 54% and cross-domain generalization by 29%, while maintaining state-of-the-art accuracy. This work establishes an interpretable and generalizable causal modeling paradigm for multi-agent trajectory prediction.

Technology Category

• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Causality
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Agentic search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Trajectory prediction models in autonomous driving are vulnerable to perturbations from non-causal agents whose actions should not affect the ego-agent's behavior. Such perturbations can lead to incorrect predictions of other agents' trajectories, potentially compromising the safety and efficiency of the ego-vehicle's decision-making process. Motivated by this challenge, we propose $ extit{Causal tRajecTory predICtion}$ $ extbf{(CRiTIC)}$, a novel model that utilizes a $ extit{Causal Discovery Network}$ to identify inter-agent causal relations over a window of past time steps. To incorporate discovered causal relationships, we propose a novel $ extit{Causal Attention Gating}$ mechanism to selectively filter information in the proposed Transformer-based architecture. We conduct extensive experiments on two autonomous driving benchmark datasets to evaluate the robustness of our model against non-causal perturbations and its generalization capacity. Our results indicate that the robustness of predictions can be improved by up to $ extbf{54%}$ without a significant detriment to prediction accuracy. Lastly, we demonstrate the superior domain generalizability of the proposed model, which achieves up to $ extbf{29%}$ improvement in cross-domain performance. These results underscore the potential of our model to enhance both robustness and generalization capacity for trajectory prediction in diverse autonomous driving domains. Further details can be found on our project page: https://ehsan-ami.github.io/critic.

Problem

• Improves robustness in autonomous driving trajectory prediction.
• Filters non-causal agent perturbations using Causal Attention Gating.
• Enhances cross-domain generalization for trajectory prediction models.

Innovation

• Causal Discovery Network identifies inter-agent causal relations
• Causal Attention Gating filters information in Transformer architecture
• Improves robustness and generalization in trajectory prediction