Study On An Multi-agent Genetic Algoiuthm Applied To The Multi-Objective Optimization Problem

Many real-world problems involve simultaneous optimization of several incommensurable and often competing objectives. These problems are formulated as Multi-objective optimization problems.This problems involve two types of problem difficulty:1)multiple, conflicting objectives and 2) a highly complex search space. On the one hand, instead of a single optimal solution competing goals give rise to a set of compromise solutions, generally denoted as Pareto-optimal. In the absence of preference information, none of the corresponding trade-offs can be said to be beter than the others. On the other hand, the search space can be too large and too complex to be solved by exact method. Conventional approaches of solving Multi-objective optimization problems include in weighting method,ε-constraint method, and objective program method and so on, but conventional approaches make a demand such as protruding, continuity, linearity to objective function, and when solving Multi- objective optimization problems, a characteristic of these conventional Multi-objective optimization algorithms is that it require turn multi- objective problem into single objective problem, however, this can always result in getting one solution of multi-objective optimization problems. Thus, efficient optimization strategies are required that are able to deal with both dificulties.In recent year,Genetic Algorithm are developed widely and paid attention gradually by people based on the thought of biology and physics. Genetic Algorithm has an advantage that it don't make a demand such as protruding, continuity,linearity to objective function. So it is applied successively to all kinds of complex optimization problems. Especialy, it is applied to Multi-objective optimization problems successively. Though GA can find the compromise solutions in limited time, improving the speed of GA is an important issue when the problem is more complex and difficult. GA poses implicit parallelism and is suitable for implementation on large scale parallel computers. Dividing the whole population into sub-populations and coarse-grained island model of exchanging information among sub-populations are the most direct parallel method.Multi-agent systems are main research areas of distributed artificial intelligence. Multi-agent systems differ from single-agent systems in that several agents exist and they are aware of each other's goals and actions. In addition to the awareness of each other's intentions and behaviour, in the fully general multi-agent system, the agents can be engaged in conversation with each other. Either to help an individual agent in achieving its goal or rarely, preventing it.In this paper, we use Genetic Algorithm and multi- agent system to solve multi-objective optimization problems.Combining the characters of the two into the Pareto concept of multi-objective optimization, a new algorithm of solving multi-objective optimization problems is presented. In this new algorithms ,both global parallelization and island parallel evolutionary algorithm models are used in a agent.Each sub-population evolves separately with different crossover and mutation probability,but they exchange best individuals in the elitist archive to increase run speed and ensure the diversity of population.We also uses the superior ordinal number method as qualification of pareto solutions.This methodnot only can get useful pareto solutions,but also avoid to turn multi- objective problem into single objective problem.To some extent, this new algorithms make up the limitation of the classical multi-objective algorithms.Numerical experiments show that the algorithms not only is simple, fast and robust but also they can get more and more extensive pareto solutions and supply more choices to decision-maker due to make use of the Pareto concept of multi-objective optimization as qualification.This new algorithm mainly has following characteristic:firstly,it uses multi-agent systems not only increasing run speed but also making it suitable for implementation on large scale parallel computers. secondly, it uses the superior ordinal number method as qualification of pareto solutions avoid to turn multi-objective problem into single objective problem.thirdly,it uses many kinds of methods to ensure the diversity of population. Finally, it uses secondary target process agent to expand the scope of multi-objective optimization problems.This algorithm similar existence many deficiencies: On the one hand, Although it has used multi-agent systems, making it suitable for implementation on large scale parallel computers.It isn't to confirm the valid of that.On the other hand,The numerical experiments is quite simple,We cann't confirm that it is also useful in a highly complex search space.author will continues to study regarding these question in the future.