DONUT: A Decoder-Only Model for Trajectory Prediction

2025-06-07

AI Summary

This paper addresses multi-agent motion prediction for autonomous driving and proposes DONUT—the first decoder-only architecture that unifies historical state encoding and future trajectory generation, departing from conventional encoder-decoder paradigms. Methodologically, it employs a decoder-only Transformer with an autoregressive iterative prediction mechanism to preserve temporal fidelity; introduces a “look-ahead prediction” strategy to model long-horizon motion trends; and integrates multi-step joint prediction (overprediction) with relative-coordinate positional encoding. Evaluated on the Argoverse 2 single-agent benchmark, DONUT achieves new state-of-the-art performance, significantly outperforming leading encoder-decoder models. These results empirically validate the effectiveness and superiority of the decoder-only paradigm for trajectory prediction, offering a novel, streamlined architectural alternative for motion forecasting in autonomous systems.

Technology Category

• Multiagent Systems: Adversarial Agents
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Search and Optimization: Learning to Search

Application Category

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

Abstract

Predicting the motion of other agents in a scene is highly relevant for autonomous driving, as it allows a self-driving car to anticipate. Inspired by the success of decoder-only models for language modeling, we propose DONUT, a Decoder-Only Network for Unrolling Trajectories. Different from existing encoder-decoder forecasting models, we encode historical trajectories and predict future trajectories with a single autoregressive model. This allows the model to make iterative predictions in a consistent manner, and ensures that the model is always provided with up-to-date information, enhancing the performance. Furthermore, inspired by multi-token prediction for language modeling, we introduce an 'overprediction' strategy that gives the network the auxiliary task of predicting trajectories at longer temporal horizons. This allows the model to better anticipate the future, and further improves the performance. With experiments, we demonstrate that our decoder-only approach outperforms the encoder-decoder baseline, and achieves new state-of-the-art results on the Argoverse 2 single-agent motion forecasting benchmark.

Problem

• Predicting agent motion for autonomous driving safety
• Replacing encoder-decoder models with decoder-only trajectory prediction
• Improving forecasting via multi-horizon overprediction strategy

Innovation

• Decoder-only model for trajectory prediction
• Single autoregressive model for encoding and predicting
• Overprediction strategy for longer horizon forecasting