Testing AI Learning in Open-World Novelty Scenarios (TALONS)

Abstract

This report transmits the results of the Option period of the Phase II continuance of the TALONS SBIR. PacMar Technologies (PMT) and Parallax Advanced Research extended the development of a set of domain independent theories, principled techniques, and software tools for generating, characterizing, and evaluating novelty. These theories and techniques covering novelties are domain independent. The agent evaluation software tool developed under this SBIR is described in this report as well. The framework developed to support the self-driving car domain (CARLA) in the base period was extended to be compatible with the Monopoly domain, Vizdoom domain, and a PMT maritime domain. Efforts under this report were to support the DARPA SAIL-ON program.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jul 14, 2023
Accession Number
AD1206086

Entities

People

  • Allison Wong
  • Bryan Grinnell
  • Dustin Dannenhauer
  • Mark Bercasio
  • Matthew Molineaux

Tags

DTIC Thesaurus Topics

  • Platforms

Readers

  • Agent-Based Social Robotics and Mobile-Assisted Learning in Virtual Environments.
  • Computer Science.
  • Computer Vision.