Ensemble of Pre-Trained Models for Long-Tailed Trajectory Prediction

2025-09-17

AI Summary

To address the long-tailed distribution challenge in autonomous vehicle trajectory prediction, this paper proposes a pre-trained large-model ensemble method that requires no retraining or fine-tuning. Our approach employs a confidence-weighted averaging strategy to fuse predictions from multiple state-of-the-art (SOTA) trajectory forecasting models. Evaluated on NuScenes and Argoverse benchmarks, it achieves a 10% improvement in overall performance over the best individual model, with consistent gains across the full data distribution—including underrepresented long-tail scenarios. To our knowledge, this is the first work to empirically demonstrate—without any additional training—that ensembling multiple SOTA models significantly improves long-tail metrics. The method is both conceptually simple and broadly applicable across diverse trajectory predictors. Implementation code is publicly available.

Technology Category

• Machine Learning: Ensemble Methods
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

This work explores the application of ensemble modeling to the multidimensional regression problem of trajectory prediction for vehicles in urban environments. As newer and bigger state-of-the-art prediction models for autonomous driving continue to emerge, an important open challenge is the problem of how to combine the strengths of these big models without the need for costly re-training. We show how, perhaps surprisingly, combining state-of-the-art deep learning models out-of-the-box (without retraining or fine-tuning) with a simple confidence-weighted average method can enhance the overall prediction. Indeed, while combining trajectory prediction models is not straightforward, this simple approach enhances performance by 10% over the best prediction model, especially in the long-tailed metrics. We show that this performance improvement holds on both the NuScenes and Argoverse datasets, and that these improvements are made across the dataset distribution. The code for our work is open source.

Problem

• Combining pre-trained trajectory prediction models without retraining
• Improving long-tailed performance in urban vehicle trajectory prediction
• Enhancing prediction accuracy through confidence-weighted ensemble methods

Innovation

• Ensemble of pre-trained models
• Confidence-weighted average method
• Out-of-box combination without retraining