Research On The Leadership For Artificial Animal Groups

Leadership in animal groups is a new problem of Artificial Life. We showan effective leadership within groups without information transferringbetween individuals, which based on the three simple steering behaviors. Insuch groups only a small proportion of informed individuals are required, butthe accuracy of groups moving is high. Furthermore we proposed a dynamicmethod for setting the coefficient in order to improve the convergence speedand accuracy, at the same time to enhance the stability of groups remarkably.Artificial Life is a new science beginning at the end of 1980s', it tries toresearch the dynamic development processes of complicated systems withcharacter of life in computers or other artificial media by using syntheticsolutions, and it extends the comprehension to the concepts of life. Theresearches of Artificial Life are helpful to discover the essential characters ofnatural lives and the basic rules of the evolution of life[1].In brief, Artificial Life is some system that its behaviors have the basiccharacters of life. In general, the basic characters of life are:self-multiplication, evolution and self-organized etc. The credendum ofArtificial Life is: life is not in isolate matter but in the combination of matter;the behaviors of life are showed in the interactive effect of massive simplematter from bottom to top. This means the matter constituting the life is notimportant. And it's the abstract property of life. We can accelerate theprocesses of researching life with the life format created easily and controlledaccurate.The more important significance may be its high abstract property forArtificial Life. Many actual complicated systems are Artificial Life systemssuch as stock system. Artificial Life let us use a new insight to look at thetraditional subjects such as economics, sociology and artificial intelligence etc.The models of opportunity format established by these subjects can't reflectthe complication character of the systems, so their results are always notaccordant to the facts or even worse. Artificial Life changes the conditionabove greatly, however.In order to research and improve the community behaviors of ArtificialLife, we have researched the constitution of individual and behaviors selectionmechanism of Artificial Life in this paper, and made some adaptivemodifications and improvements, then established the platform for researchthe virtual community movements of animals. On this platform weimplemented the simulations and emulation in computer of the animalindividual related to other existent creatures in nature, and displayed the basiccharacters of the creatures such as behavior, individual and community, andself-organized etc. The animal individual consists of vision system, movementsystem, behavior selection system and state system. Behavior selection systemis like the brain of an animal, it gets the information of environment, deliveredamong individuals and state collected by vision system, and uses somemechanism should be used currently and controls the implement according tothe information. In this paper the "simulating-driving" mechanism and the"motivation-based" mechanism is combined to use, which makes the animalshave the ability of harmony in many complicated behaviors. The animalsconsider both their condition and the effect of the factors outside when theyselect which behavior. Like "simulating-driving", firstly we select thebehavior having the biggest(best) value from all options, then make the restbehaviors unselected affect the selected one, which can make the animalscompromise in many methods.There are two kinds of community movement leading mechanism: simplecommunity movement leading mechanism and distributed communitymovement leading mechanism. The theory of the former is: a leader is addedinto the community, and it is supposed to have much information, specificdirection and ability of deciding the behavior should be taken according to thecondition of the whole community in every case. The other individuals obeythe leader's command and move in the state selected by the leader. So, it hasobvious disadvantage: the movements of community depend on the leadingexcessively and need to deliver the information, which makes them blockexcessively and reduces the efficiency of the system.The current distributed community movement leading mechanism isbased on three pieces of basic movement law. It ensures the stability of thecommunity greatly, meaning it is not easy to separate from community forindividual, and needn't to deliver any kind of information among theindividuals. But there is a fatal disadvantage in it: the directions of themovements of community have no specific objective because they varyrandom and constantly, which is different from the ones of the most animals innature obviously.In this paper, we proposed the improved distributed communitymovement leading mechanism and a dynamic method for setting thecoefficient in order to implement the community movements, correspond tothe facts, maintain the reliability and stability of community and make thecommunity has specific movement directions and regular intervals.There is no leader in this mechanism, there are a few "knower" havingcertain experience or the information of the objective;the other individualsare "un-knower" having no any information. It is no need to deliver any kindof information among the animals in community, so, an animal individualdoesn't know which is "knower" or "un-knower" among its companies in therest of community, and doesn't know which "knower" has the rightinformation. The movement directions of the individual who is "un-knower"obey three pieces of basic movement law, while the individual who is"knower" has the anticipant direction self-controlled besides the basic law andit can compromise between the two options through coefficientω .All individuals in community obey or consult three pieces of basicmovement law, which makes the stability and direction consistency of thewhole community even though there are dispersive movements of individual.The knower adds an anticipant direction self-controlled into the optionaldirections. They will lead the whole community to move towards the rightrelative direction, which ensures the accuracy of the direction of the wholecommunity.ω is the coefficient which shows the rate of the anticipant directionscompared to the ultimate directions of knower, it acts an important function inthe whole process. The anticipant directions and the basic directions arecollaborative and conflictive mutually when knower selects the directions.The value of ω is the standard to compromise appropriately between the twoconflictive options: the bigger value, the faster speed and more accurate of thecommunity movements trending to the right direction, but meanwhile, thesmaller rate of basic movement law compared to the movement directions ofknower, and the higher probability of separating;the smaller value, theopposite cases. So a dynamic method for setting the coefficient is used in thispaper: every individual who is knower calculates the angle between itsdirection and the community' when it selects the direction, then decides thevariety of ω according to the angle.At last, we have done a lot of experiments on the improved distributedcommunity movement leading mechanism and the dynamic method for settingthe coefficient, and obtained perfect results, which proved the validity of theimproved distributed community movement leading mechanism.