Summaries - Office of Research & Innovation
Research Summaries
Back Leverage Artificial Intelligence (AI) to Learn, Optimize, and Win (LAILOW) for a Complex Enterprise - Navy Logistics and Supply Chain
Fiscal Year | 2020 |
Division | Graduate School of Operational & Information Sciences |
Department | Information Sciences |
Investigator(s) | Zhao, Ying |
Sponsor | Office of Naval Research (Navy) |
Summary |
A complex enterprise with multiple subsystems and organizations is omnipresent. A Navy fleet is a complex enterprise and a Navy logistics and supply chain is a complex enterprise as well. Even a Naval maintenance and repair operation for a specific product such MV-22 is a complex enterprise. A complex enterprise contains a myriad of business processes as subsystems can be either sequential or parallel. A complex enterprise needs trusted AI to achieve automation, foster collaborations, and win competitions. How we can leverage data sciences and advanced AI techniques for complex enterprises is the topic of this proposal. Each process in a complex enterprise has a narrative on what is designed to achieve for the whole enterprise. We propose to demonstrate an innovative framework that leverages artificial Intelligence (AI) to learn, optimize, and win (LAILOW) for a complex Navy enterprise. The long-term objective of the project is to architect and demonstrate LAILOW for a resilient and agile logistics of the U.S. Navy to improve readiness and sustainment at peace time or at war (under persistent multi-domain attack) by learning, optimizing, and gaming the performance of the functions of total logistics enterprise including supply, maintenance, transportation, health services, general engineering, and finance. For the proposed performance period, we will demonstrate the feasibility of LAILOW by addressing two of the key challenges for U.S. Marine Corp. (USMC) supply chain and logistics: 1) there is need to apply holistic ML/AI approaches to improve the total USMC combat readiness; 2) there is uncertainty, no data situations for deep analytics and ML/AI models which might require large-scale simulations and what if analyses. We will address the two challenges by focusing on one business problem that the current Marine Corps logistics information systems do not possess predictive modeling & simulation tools that can integrate with intelligence data and other subject matter experts (SMEs) input to aid in logistical planning efforts to support a Marine Air-Ground Task Force (MAGTF). This problem is especially challenging because such a unit's structure (table of organization and equipment) may include a large of parts, each one has specific need for the (X) duration and frequency of manpower and equipment for maintenance. Based on history data, a predictive model in this domain needs to compute the most probable parts (failure rates/demand history/available manpower) to support the unit's operation in normal condition. Analysis of Alternatives (AoA) will be conducted as simulations and what if analysis to create added conditions such as an IED blast, desert environment, and corrosion considerations. In order to compute "delta" for replenish based on the perturbations to the complex system, gathered intelligence data, knowledge and rules from SMEs, and existing predictive and engineering models will feed to the system to provide extra data to predict such "delta." It is important to perform this type of AoA or simulations since there might be no historical data available for new conditions, traditional predictive modeling analysis might not be directly applicable. We propose to look at the simulation data to modify the demand model based on machine learning such as reinforcement learning, unsupervised learning, and graph theory. For example, Soar reinforcement learning (Soar-RL) to learn and modify the existing knowledge rules from external environmental reward and penalty for a complex enterprise. The unsupervised learning Lexical Link Analysis (LLA) discovers patterns and rules from historical or simulation data into networks and graphs. Graph theory can then be applied to provide a global, holistic, and associated view of spare parts needed for when multiple new conditions that occur as perturbations to the complex enterprise. |
Keywords | |
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 |