Skip to content

Modeling Subjectivity in Trajectories

Home / Open Questions / Modeling Subjectivity in Trajectories

Background: Trajectory prediction models often struggle to explicitly learn and incorporate the subjective, anisotropic behaviors that influence individual agent movement.

Question / Future Work: Developing more robust methods to explicitly represent, learn, and modulate trajectory predictions based on the unique, time-varying subjective intentions or ‘biases’ of agents remains a significant challenge for human-like social navigation. Future work is needed to refine the integration of these inferred subjectivities to ensure they provide stable and high-fidelity conditioning for complex multi-agent forecasting tasks.

Why It Matters: This represents a fundamental bottleneck in achieving human-like predictability in complex social navigation scenarios, where agent subjectivity is a primary driver of anisotropic behavior.

Evidence: Despite great efforts, it remains difficult to explicitly learn and forecast these subjectivities, let alone further modulate models’ predictions through a specific ego’s subjectivity.

Metadata & Links