LongComp: Long-Tail Compositional Zero-Shot Generalization for Robust Trajectory Prediction

2025-11-13

AI Summary

Autonomous driving trajectory prediction must address rare yet safety-critical long-tail scenarios that are underrepresented in real-world data. To tackle this, we propose the first robust evaluation framework for compositional zero-shot generalization: it decouples ego-vehicle behavior from social interaction context to construct challenging out-of-distribution (OOD) test sets; introduces safety-aware scene factor decomposition, a task-modularized gating network, and a difficulty-prediction auxiliary head—jointly enhancing both in-distribution (ID) and OOD performance. Under closed-set and open-set compositional generalization settings, our method reduces the OOD performance gap by up to 2.8% and 11.5%, respectively, while simultaneously improving ID prediction accuracy. This work pioneers the integration of modular gating and difficulty-aware mechanisms into trajectory prediction robustness modeling, establishing a verifiable evaluation and optimization paradigm for safety-critical long-tail scenarios.

Technology Category

• Humans and AI: Human-Aware Planning and Behavior Prediction
• Natural Language Processing: Safety and Robustness
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Security and Privacy: Large-scale security measurements
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI

Abstract

Methods for trajectory prediction in Autonomous Driving must contend with rare, safety-critical scenarios that make reliance on real-world data collection alone infeasible. To assess robustness under such conditions, we propose new long-tail evaluation settings that repartition datasets to create challenging out-of-distribution (OOD) test sets. We first introduce a safety-informed scenario factorization framework, which disentangles scenarios into discrete ego and social contexts. Building on analogies to compositional zero-shot image-labeling in Computer Vision, we then hold out novel context combinations to construct challenging closed-world and open-world settings. This process induces OOD performance gaps in future motion prediction of 5.0% and 14.7% in closed-world and open-world settings, respectively, relative to in-distribution performance for a state-of-the-art baseline. To improve generalization, we extend task-modular gating networks to operate within trajectory prediction models, and develop an auxiliary, difficulty-prediction head to refine internal representations. Our strategies jointly reduce the OOD performance gaps to 2.8% and 11.5% in the two settings, respectively, while still improving in-distribution performance.

Problem

• Addresses rare safety-critical scenarios in autonomous driving trajectory prediction
• Creates challenging out-of-distribution test sets through dataset repartitioning
• Improves generalization using modular networks and difficulty-prediction methods

Innovation

• Safety-informed scenario factorization disentangles contexts
• Task-modular gating networks enhance trajectory prediction
• Auxiliary difficulty-prediction head refines internal representations