Online Aggregation of Trajectory Predictors

2025-02-11

AI Summary

Trajectory prediction models for autonomous driving suffer significant performance degradation under out-of-distribution (OOD) environmental conditions. Method: This paper proposes a lightweight, model-agnostic online ensemble method that dynamically fuses outputs from multiple heterogeneous predictors. It is the first to introduce non-convex, non-stationary online optimization into trajectory prediction ensembling—requiring no environmental priors or model retraining, and relying solely on real-world trajectory feedback to drive adaptive weight updates. The approach extends the online convex optimization framework by designing expert-weighted probability vectors and an adaptive loss function to enable efficient gradient-based updates. Contribution/Results: When trained on NuScenes and transferred cross-domain to the Lyft dataset, the proposed aggregator matches or surpasses the performance of the single best individual model, demonstrating substantial improvements in robustness and generalization under distributional shift.

Technology Category

• Search and Optimization: Learning to Search
• Machine Learning: Ensemble Methods
• Multiagent Systems: Adversarial Agents

Application Category

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

Abstract

Trajectory prediction, the task of forecasting future agent behavior from past data, is central to safe and efficient autonomous driving. A diverse set of methods (e.g., rule-based or learned with different architectures and datasets) have been proposed, yet it is often the case that the performance of these methods is sensitive to the deployment environment (e.g., how well the design rules model the environment, or how accurately the test data match the training data). Building upon the principled theory of online convex optimization but also going beyond convexity and stationarity, we present a lightweight and model-agnostic method to aggregate different trajectory predictors online. We propose treating each individual trajectory predictor as an"expert"and maintaining a probability vector to mix the outputs of different experts. Then, the key technical approach lies in leveraging online data -the true agent behavior to be revealed at the next timestep- to form a convex-or-nonconvex, stationary-or-dynamic loss function whose gradient steers the probability vector towards choosing the best mixture of experts. We instantiate this method to aggregate trajectory predictors trained on different cities in the NUSCENES dataset and show that it performs just as well, if not better than, any singular model, even when deployed on the out-of-distribution LYFT dataset.

Problem

• Aggregate diverse trajectory predictors online
• Leverage online data to optimize expert mixture
• Improve prediction accuracy across different datasets

Innovation

• Online aggregation of predictors
• Model-agnostic expert mixing
• Dynamic loss function optimization