Introducing Competition in a Multi-Agent System for Hybrid Optimization

Process systems engineering optimization problems may be challenging. These problems often exhibit nonlinearity, non-convexity, discontinuity, and uncertainty, and often only the values of objective and constraint functions are accessible. Additionally, some problems may be computationally expensive. In such scenarios, black-box optimization methods may be appropriate to tackle such problems. A general-purpose multi-agent framework for optimization has been developed to automate the configuration and use of hybrid optimization, allowing for multiple optimization solvers, including different instances of the same solver. Solvers can share solutions, leading to better outcomes with the same computational effort. Alongside cooperation, competition is introduced by dynamically allocating more computational resource to solvers best suited to the problem. Each solver is assigned a priority that adapts to the evolution of the search. The scheduler is priority-based and uses similar algorithms to those in multi-tasking operating systems. The impact on the number of priority levels is investigated. The framework allows for the use of both metaheuristic and direct search methods. Metaheuristics explore the full search space while direct search methods are good at exploiting solutions. The framework has been implemented in Julia, making full use of multiprocessing. A multiobjective case study on the design of a micro-analytic system is presented. The case study demonstrates the benefits of a multi-solver hybrid optimization approach with both cooperation and competition. The framework adapts to the evolving requirements of the search. Often, a metaheuristic method is allocated more computational resource at the beginning of the search while direct search methods are prioritized later.