Modeling Conditional Dependencies in Multiagent Trajectories

Research output: Journal contributionsConference article in journalResearchpeer-review

Authors

We study modeling joint densities over sets of random variables (next-step movements of multiple agents) which are conditioned on aligned observations (past trajectories). For this setting, we propose an autoregressive approach to model intra-timestep dependencies, where distributions over joint movements are represented by autoregressive factorizations. In our approach, factors are randomly ordered and estimated with a graph neural network to account for permutation equivariance, while a recurrent neural network encodes past trajectories. We further propose a conditional two-stream attention mechanism, to allow for efficient training of random factorizations. We experiment on trajectory data from professional soccer matches and find that we model low frequency trajectories better than variational approaches.
Original languageEnglish
JournalProceedings of Machine Learning Research
Volume151
Pages (from-to)10518-10533
Number of pages16
ISSN2640-3498
Publication statusPublished - 2022
EventThe 25th International Conference on Artificial Intelligence and Statistics - AISTATS 2022 - Virtual Conference, Online
Duration: 28.03.202230.03.2022
Conference number: 25
http://aistats.org/aistats2022/

Bibliographical note

Funding Information:
We would like to thank Marius Lehne at SAP SE for his essential support and help while writing this paper as well as all reviewers for their valuable feedback.

Publisher Copyright:
Copyright © 2022 by the author(s)

Recently viewed

Publications

  1. Multi-view learning with dependent views
  2. Robust decoupling through algebraic output feedback in manipulation systems
  3. Changing the Administration from within:
  4. Modeling items for text comprehension assessment using confirmatory factor analysis
  5. Age effects on controlling tools with sensorimotor transformations
  6. Development of a scoring parameter to characterize data quality of centroids in high-resolution mass spectra
  7. A Comparative Study for Fisheye Image Classification
  8. Integration of demand forecasts in ABC-XYZ analysis
  9. Assessment of university students’ understanding of abstract binary operations
  10. Comparison of EKF and TSO for Health Monitoring of a Textile-Based Heater Structure and its Control
  11. Improving mechanical properties of chip-based aluminum extrudates by integrated extrusion and equal channel angular pressing (iECAP)
  12. Accuracy and bias of methods used for root length measurements in functional root research
  13. A Stacked Planar Sensor Concept for Minimally Invasive Plasma Monitoring
  14. General Patterns and Conclusions
  15. "Learning by doing"
  16. The uses of isospin in early nuclear and particle physics
  17. Developing pragmatic competence using EFL textbooks: Focus on requests
  18. Explaining the (Non-) Adoption of Advanced Data Analytics in Auditing
  19. Green your community click by click
  20. A Note on Pensions and Firm Performance
  21. Reframing the possible
  22. Revisiting Carbon Disclosure and Performance
  23. Visual-Inertial Navigation Systems and Technologies
  24. Value of large-scale linear networks for bird conservation
  25. Competition between honey bees and wild bees and the role of nesting resources in a nature reserve
  26. Flat-of-the-curve medicine
  27. Combination matters
  28. Ambient Assited Energy Management