Heuristics for Solver-Aware Systems Architecting (SASA): A Reinforcement Learning Approach

Abstract

Abstract The crowdsourcing literature has shown that domain experts are not always the best solvers for complex system design problems. Novices and specialists in adjacent domains can, under certain conditions, provide novel solutions at lower costs. Additionally, the best types of solvers for different sub-problems are dependent on the architecture of complex systems. The assignment of solvers based on the architecture, referred to as Solver-Aware System Architecting (SASA), expands traditional system architecting practices by considering solver characteristics and contractual incentive mechanisms in the design process and aims to improve complex system design and innovation by leveraging the strengths of domain experts, crowds, and specialists for different parts of the problem. Given the complexity of system design problems and the variety of solvers available, it is desirable to have heuristics to guide solver assignment to different problems. Developing effective heuristics for solver assignments in complex system design is challenging due to a large number of possible combinations of problem-solver pairs. To address this challenge, this paper presents a computational approach using a multi-armed bandit (MAB) formulation to generate heuristics for solver assignment. The approach is demonstrated using a simple and idealized problem of golf, which has characteristics similar to design problems, including how the problem is decomposed into sub-problems, and solved by different solvers. The results show that the proposed approach is effective in deriving a rich set of heuristics for the golf problem, and can be extended in the future to more complex systems design problems.