Researchers at MegaOne AI, led by Senior Staff Scientist Dr. Anya Sharma, have introduced AutoMS, a novel multi-agent evolutionary search framework designed for cross-physics inverse microstructure design. The work, detailed in a paper published on arXiv on March 27, 2026, addresses the complex challenge of designing microstructures that simultaneously meet multiple, often conflicting, physical objectives across different domains, such as mechanical and thermal properties. This research aims to accelerate the discovery of advanced materials with tailored functionalities by navigating the vast and discontinuous search spaces inherent in inverse design problems. The full paper is available at arXiv:2603.27195.
AutoMS employs a multi-agent system where each agent specializes in optimizing a specific physical objective. These agents collaborate and compete within an evolutionary search framework, allowing for the simultaneous exploration and exploitation of the design space. The system integrates a surrogate modeling approach to reduce the computational cost associated with high-fidelity physics simulations, a critical bottleneck in traditional inverse design methods.
The framework was tested on several benchmark problems involving coupled mechanical and thermal properties. In one experiment, AutoMS successfully designed microstructures exhibiting a target Young’s modulus of 10 GPa and a thermal conductivity of 50 W/(m·K), achieving a Pareto front that outperformed single-objective optimization approaches by an average of 15% in terms of objective space coverage. This demonstrates its capability to identify designs that balance competing physical requirements.
A key technical detail of AutoMS is its use of a genetic algorithm coupled with a deep neural network-based surrogate model. This surrogate model was trained on a dataset of 10,000 microstructure-property pairs, enabling rapid prediction of physical properties with an R-squared value of 0.97 for both mechanical and thermal responses, significantly reducing the need for repeated finite element analyses during the evolutionary search process.
Furthermore, the multi-agent architecture incorporates a dynamic weighting scheme that adjusts the influence of each agent based on the current state of the population and the proximity to the desired multi-objective targets. This adaptive mechanism helps guide the search towards optimal trade-offs, particularly in highly constrained design problems. The system demonstrated a 3x speedup in convergence to a satisfactory Pareto front compared to a baseline multi-objective evolutionary algorithm without surrogate modeling.
The current iteration of AutoMS primarily focuses on 2D microstructure design problems. Future work will involve extending the framework to handle complex 3D microstructures and incorporating additional physics domains, such as electrical or acoustic properties, to further broaden its applicability in materials science and engineering.
