A Simulation Framework of Multi-objective Evolutionary Algorithms and Surrogate-based Optimization for Guided Weapon Design

The goal of the proposed effort is to develop a simulation architecture based on multi-objective evolutionary algorithms, surrogate models, and hardware acceleration for optimization and design of multi-disciplinary autonomous munition systems. In Phase I, a modular simulation framework encapsulating key technology elements, including high-fidelity guided weapon simulator, multi-objective evolutionary algorithms (MEA), surrogate-based optimization, and GPU computing was successfully developed. Proof-of-concept was demonstrated via USAF relevant case studies that the new paradigm enables unprecedented computational performance for guided weapon design. The design identified by our tool always outperform the benchmark. The surrogate-based optimization enables 61.3% reduction in computational time, and GPU computing exhibits 100ï‚´ speedup over CPU in model evaluation. In Phase II, the framework will be enhanced in performance and functionality. Guided weapon modeling will be extended using USAF simulators to capture design complexity and fidelity. MEA and surrogate models will be improved to accommodate broad and deep design space in practical munition systems and evolving targets. Development of scalable software along with ease-to-use user interface and run-time visualization on heterogeneous CPU-GPU computing platform to exploit hardware advancements. The technology will be extensively validated and demonstrated using case studies of USAF interest, and integrated with USAF designated guided munition design