A. Grosvenor
MSBAI,
United States
Keywords: Simulation, Computer-Aided Engineering, Electric Vehicles
Summary:
Two fundamental problems exist to simulation use: Learning one simulation package takes an engineer typically months to years — and there are more than 1000 such packages covering a broad range of applications, and setting up a single new simulation commonly takes hours of manual work. Only single digit percentages of engineers use simulation as a tool in their design toolbox; MSBAI will make it 100%. MSBAI develops a 'Universal Interface for Simulation' called GURU — a cognitive AI assistant that enables untrained users to run specialized simulation software. GURU is an 'AI layer' between human and computer, and will become a new commercial solution to enabling engineers to begin using a broad range of powerful simulation software like never before. GURU is a highly modular agent-based, blackboard-driven system. At the platform level, the goal manager interprets inputs from the user (‘intents’ via the UI) and assembles workflows consisting of skills/capabilities trained in agents. We have chosen a common hybrid AI architecture for each agent. The hybrid of symbolic learning and geometric/machine learning enables the system to perform trusted explainable procedures, while sufficiently generalizing to real-world requirements. Our learning engine, ARTIST (Automated Reinforcement-learning-guided Tree-based hybrid-Intelligence Synthesis Trainer), employs a hybrid of: ‘Program exploration’ - building state and feature graphs, training transformer-based Q-value estimation, and a history-aware Reinforcement Learning approach. GURU addresses three simulation market verticals: computer-aided engineering, trajectory and mission planning, and virtual world immersive training. MSBAI’s been awarded two SBIR Phase II contracts: 1) to adapt GURU to reduce the time needed to set up new threat trajectory simulations for the USAF and MDA, 2) to adapt GURU, radically scaling up utilization of engineering simulation to address the USAF's need to reduce the time to design new aerial platforms. Last fall the USAF announced a digital transformation, where simulation based engineering design would be scaled up to a level that is unprecedented, to reduce the development time of new aerial platforms down from a decade+ to one single year. Then USAF Acquisition Chief, Dr. Will Roper, adopted the term ‘e-system’ to denote virtual prototypes designed, evolved, and tested in simulation first. MSBAI’s Phase II SBIR contract in the 20.3 cycle addresses this challenge. The same role of simulation in accelerating critical engineering innovation identified by the USAF can clearly be applied to the demand in the EV race in order to accelerate vehicle development. The EV market is projected to be $5 trillion in the next decade, and China currently has 3x the EV sales as the US. In 2021 the White House said, “it is time for ...the U.S. to lead in EV manufacturing, infrastructure, deployment, and innovation.” GURU will allow the autonomous setup and execution of engineering simulations and is a solution to the urgent digital engineering needs of the automotive industry, allowing the US to dominate the EV race.