User Tools

Site Tools

Action unknown: copypageplugin__copy

9a.016.UVA - Data Driven Approach to Jointly Modeling Intent for Navigating Diverse Agents

Project - Summary

  • Most autonomous agents are required to navigate safely in dense, interacting environments. Safe navigation can be challenging, especially when the environment consists of heterogeneous traffic, i.e. agents with diverse shapes, trajectory dynamics and expected behavior (e.g. cars, cyclists, pedestrians, etc.).
  • Further, safe navigation for an agent can potentially be achieved by multiple socially acceptable (yielding right-of-way, respecting personal space, etc.) and physically feasible (collision-free, feasible terrain, etc.) future trajectories.
  • In this project, we aim to develop a data-driven approach to multi-agent intent prediction, that can generate multiple socially acceptable and physically plausible future trajectories for co-navigating heterogeneous agents in a unified framework.

Project - Team

Team Member Role Email Phone Number Academic Site/IAB
Peter Beling PI (804) 982-2066 University of Virginia

Project - Novelty of Approach

- Our existing work was developed with particular focus on maritime domain and evaluated on ship trajectory datasets. In this project, we aim to extend the work to autonomous systems in other domains, e.g., self-driving cars, unmanned aerial vehicles, etc.
- Currently, our framework is able to model correlations between trajectories of agents with nearly similar dynamics (e.g. different kinds of marine vessels). Realistically, in complex environments such as roads, different kinds of agents co-navigate, e.g., cars, pedestrians, cyclists, etc. In this project, we aim to model correlations between such diverse trajectory dynamics and take those into account while generating trajectory predictions.
- Unlike our existing work, this project aims at generating multiple socially acceptable and physically feasible trajectory predictions for agents in a scene.

Project - Deliverables

1 Briefing of model design
2 Literature review document
3 Final report

Project - Benefits to IAB

This project will directly inform technology development for autonomous systems. It will benefit any IAB member that is developing autonomous agents.

Project - Documents

projects/year9/9a.016.uva.txt · Last modified: 2021/05/20 10:30 by sally.johnson