Summaries - Office of Research & Innovation
Research Summaries
Back Design, Demonstrate and Proof of Concept of Using the Explainable Reinforcement Learning (xRL) in Soar for Combat Identification (CID)
Fiscal Year | 2019 |
Division | Graduate School of Operational & Information Sciences |
Department | Information Sciences |
Investigator(s) | Zhao, Ying |
Sponsor | Defense Advanced Research Projects Agency (DoD) |
Summary | There has been tremendous advancement in commercial applications recently using the Big Data, Deep Analytics including machine learning (ML) and artificial intelligence (AI) methods. These methods also provide emerging technologies to address the unique challenges in the Common Tactical Air Pictures (CTAP) and Combat Identification (CID) applications. It is also imperative to test and adapt commercially available tools now to meet the ongoing needs and requirements for the Common Tactical Air Pictures (CTAP) and Combat Identification (CID) requirements. In the past, we have been researching and testing on the Soar (Laird, 2012) cognitive architecture including reinforcement learning (RL) algorithm. One of the Soar advantages to start for the CTAP and CID application is that it is able to easily incorporate existing English like production rules. We also use a data-driven lexica link analysis (LLA) to initialize/discover new rules for the Soar-RL. The explainable characteristics of Soar-RL is a very important feature for CID, where the decision rules are presented for each reasoning. In this project, we will work the DARPA XAI community to design, improve and implement the XAI feature of the Soar-RL for a better human-machine interface, explanation and cognitive architecture that address and improve the cognitive CID. |
Keywords | Artificial Intelligence Reinforcement Learning XAI combat identification common tactical air pictures xplainable AI |
Publications | Publications, theses (not shown) and data repositories will be added to the portal record when information is available in FAIRS and brought back to the portal |
Data | Publications, theses (not shown) and data repositories will be added to the portal record when information is available in FAIRS and brought back to the portal |