Dynamic Intent Queries for Motion Transformer-based Trajectory Prediction

2025-04-22

AI Summary

Motion Transformer (MTR) suffers from trajectory–map topological inconsistency and physically infeasible predictions due to its reliance on predefined static intent points. Method: We propose a scene-aware dynamic intent query mechanism that jointly encodes map topology and agent interaction features, generating traffic-scenario-adaptive dynamic intent points via a learnable module—replacing fixed prior points. Contribution/Results: This work is the first to embed dynamic, scene-driven intent generation into the MTR architecture, overcoming inherent limitations of static intent modeling. Experiments on the Waymo Open Motion Dataset demonstrate significant improvements: reduced average displacement error (ADE) and final displacement error (FDE) for 5-second long-horizon prediction; enhanced robustness against illegal or map-violating ground-truth trajectories; and substantially improved map compliance and physical plausibility of predicted trajectories.

Technology Category

• Intelligent Robots: Motion and Path Planning
• Planning, Routing, and Scheduling: Model-Based Reasoning
• 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: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

In autonomous driving, accurately predicting the movements of other traffic participants is crucial, as it significantly influences a vehicle's planning processes. Modern trajectory prediction models strive to interpret complex patterns and dependencies from agent and map data. The Motion Transformer (MTR) architecture and subsequent work define the most accurate methods in common benchmarks such as the Waymo Open Motion Benchmark. The MTR model employs pre-generated static intention points as initial goal points for trajectory prediction. However, the static nature of these points frequently leads to misalignment with map data in specific traffic scenarios, resulting in unfeasible or unrealistic goal points. Our research addresses this limitation by integrating scene-specific dynamic intention points into the MTR model. This adaptation of the MTR model was trained and evaluated on the Waymo Open Motion Dataset. Our findings demonstrate that incorporating dynamic intention points has a significant positive impact on trajectory prediction accuracy, especially for predictions over long time horizons. Furthermore, we analyze the impact on ground truth trajectories which are not compliant with the map data or are illegal maneuvers.

Problem

• Static intention points misalign with map data in traffic scenarios
• Dynamic intention points improve trajectory prediction accuracy
• Addressing unrealistic goal points in Motion Transformer models

Innovation

• Dynamic intention points replace static ones
• Enhanced MTR model for trajectory prediction
• Improved accuracy in long-term predictions