(SinBaD) Simulation in Behaviors and Decisions
Navy SBIR FY2018.1


Sol No.: Navy SBIR FY2018.1
Topic No.: N181-083
Topic Title: (SinBaD) Simulation in Behaviors and Decisions
Proposal No.: N181-083-1127
Firm: Knowledge Based Systems, Inc.
1408 University Drive East
College Station, Texas 77840
Contact: Tim Darr
Phone: (979) 260-5274
Web Site: http://www.kbsi.com
Abstract: Knowledge-Based Systems Incorporated (KBSI) proposes �?oSimulation in Behaviors and Decisions�?? (SinBaD) in response to the Navy SBIR �?oWarfighting Chess Games and Pieces�?? (topic number N181-083). We will demonstrate proof-of-concept simulation components that have the capability to pit smart red/gray forces against smart blue forces to aid development of timely and effective decision support tools. This proposed capability will provide a service that is available for decision support tools to suggest modifications to current plans and predict future outcomes contextualized to a specific situation. Our demonstration scenario will be a lab-based 3-on-3 game in which both the red/gray commander and blue commander have three platoons with an objective of securing an area of interest. To increase our understanding, we will vary the parameters of the game to explore the differences in strategies. For example, the differences between an amphibious operation to secure a beachhead as opposed to one in a dense, urban environment to secure critical infrastructure. The Phase I results will demonstrate the feasibility of the approach for an operationally relevant toolset and decision support instrument. The Phase I results will also identify metrics to verify performance, and model normalization across simulation systems necessary to reduce technical risk.
Benefits: The expected SiNBaD benefits include �?� Agent behavior models derived from the behavioral economics discipline produce more accurate strategic decision making within a simulation �?� Decision making �?oerrors�?? modeled on real human behaviors ensure fair fights in agent vs. agent simulation �?� An unsupervised deep learning method (generative adversarial networks) learns decision making styles of specific behavioral profiles �?� Thousands of agent vs. agent simulations can be performed in a short period of time to aid development of effective decision support tools

Return