It consists of a society of N agents where they cooperate in a sharing environment (E) to realize a common intention of finding high-quality solution(s) for well-defined landscape (G) of the optimization task.

Each agent in the MAOS has a moderate problem solving capability with two basic features: a) addresses on possessing of the long-term declarative memory in limited capacity, which is modified reactively by itself only, instead of operating as sophisticated as in cognitive architectures; and b) achieves complex problem solving by simple guided generate-and-test behavior under the law of socially biased individual learning (SBIL), a fast-and-frugal heuristic under bounded rationality adopted by many species in the real world.

Some algorithms, including Particle Swarm Optimization (PSO), Differential Evolution (DE), Social Cognitive Optimization (SCO), and Electromagnetism-like Mechanism (EM) Heuristic, etc, for Numerical Optimization Problems (NOP) have been implemented in the SWAF. A simplified example of the MAOS is the Mini-Swarm System, which has been applied on some combinatorial optimization problems, such as Traveling Salesman Problem (TSP), Qudratic Knapsack Problem (QKP), etc.

Related Papers