Flexible Multi-Agent Trajectory Modeling
Background: Multi-modal generative models for trajectory forecasting and completion often operate on fixed temporal horizons, which limits their adaptability to diverse real-world task settings and scenarios.
Question / Future Work: Developing multi-agent trajectory modeling architectures that are not constrained by fixed temporal window dimensions or a predefined number of agents remains a challenging open problem for ensuring broader applicability.
Why It Matters: Flexibility in temporal and spatial dimensions is critical for applying trajectory modeling to varying, real-world sports and surveillance scenarios where observation durations and agent counts change dynamically.
Metadata & Links
- created_at
- 2026-05-14T06:10:38Z