Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Enhanced Prediction of Multi-Agent Trajectories via Control Inference and State-Space Dynamics

Yu Zhang, Yongxiang Zou, Haoyu Zhang, Zeyu Liu, Houcheng Li, Long Cheng

TL;DR

This paper addresses multi-agent trajectory prediction for autonomous systems by integrating state-space dynamics with graph neural networks. It introduces a novel Mixed Mamba encoder to infer initial control variables, pairs a control-variable evolution modeled as a state-space system, and uses an Extended Kalman Filter to provide uncertainty estimates. The framework leverages a graph-aware encoder to preserve inter-agent interactions and demonstrates superior performance on public datasets (highD and rounD) across multiple metrics, with ablation studies highlighting the value of graph information and mixed Mamba components. The work advances interpretable, physically grounded trajectory forecasting at scale, with implications for safer and more efficient autonomous navigation.

Abstract

In the field of autonomous systems, accurately predicting the trajectories of nearby vehicles and pedestrians is crucial for ensuring both safety and operational efficiency. This paper introduces a novel methodology for trajectory forecasting based on state-space dynamic system modeling, which endows agents with models that have tangible physical implications. To enhance the precision of state estimations within the dynamic system, the paper also presents a novel modeling technique for control variables. This technique utilizes a newly introduced model, termed "Mixed Mamba," to derive initial control states, thereby improving the predictive accuracy of these variables. Moverover, the proposed approach ingeniously integrates graph neural networks with state-space models, effectively capturing the complexities of multi-agent interactions. This combination provides a robust and scalable framework for forecasting multi-agent trajectories across a range of scenarios. Comprehensive evaluations demonstrate that this model outperforms several established benchmarks across various metrics and datasets, highlighting its significant potential to advance trajectory forecasting in autonomous systems.

Enhanced Prediction of Multi-Agent Trajectories via Control Inference and State-Space Dynamics

TL;DR

This paper addresses multi-agent trajectory prediction for autonomous systems by integrating state-space dynamics with graph neural networks. It introduces a novel Mixed Mamba encoder to infer initial control variables, pairs a control-variable evolution modeled as a state-space system, and uses an Extended Kalman Filter to provide uncertainty estimates. The framework leverages a graph-aware encoder to preserve inter-agent interactions and demonstrates superior performance on public datasets (highD and rounD) across multiple metrics, with ablation studies highlighting the value of graph information and mixed Mamba components. The work advances interpretable, physically grounded trajectory forecasting at scale, with implications for safer and more efficient autonomous navigation.

Abstract

In the field of autonomous systems, accurately predicting the trajectories of nearby vehicles and pedestrians is crucial for ensuring both safety and operational efficiency. This paper introduces a novel methodology for trajectory forecasting based on state-space dynamic system modeling, which endows agents with models that have tangible physical implications. To enhance the precision of state estimations within the dynamic system, the paper also presents a novel modeling technique for control variables. This technique utilizes a newly introduced model, termed "Mixed Mamba," to derive initial control states, thereby improving the predictive accuracy of these variables. Moverover, the proposed approach ingeniously integrates graph neural networks with state-space models, effectively capturing the complexities of multi-agent interactions. This combination provides a robust and scalable framework for forecasting multi-agent trajectories across a range of scenarios. Comprehensive evaluations demonstrate that this model outperforms several established benchmarks across various metrics and datasets, highlighting its significant potential to advance trajectory forecasting in autonomous systems.
Paper Structure (17 sections, 18 equations, 4 figures, 2 tables)

This paper contains 17 sections, 18 equations, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. RELATED WORK
  3. Traffic Behavior Prediction
  4. GNNs for Trajectory Prediction
  5. Applications of State Space Models
  6. Preliminaries
  7. Model engineering
  8. Data-driven system identification
  9. Black-box model learning
  10. The Proposed Model
  11. Dynamic modeling
  12. Graph mixed Mamba encoder
  13. Control variable SSM
  14. EKF
  15. Evaluation and Results
  16. ...and 2 more sections

Figures (4)

  • Figure 1: The main difference between the proposed framework and traditional frameworks for multi-agent trajectory prediction.
  • Figure 2: The overall structure of the proposed algorithm.
  • Figure 3: The overall structure of the proposed mixed Mamba.
  • Figure 4: The designed structure of the evolving control variables.