Research on Ant Colony Algorithm

Follow the trail of ants. What did you find? Through the above principle description and practical operation, it is not difficult to find that the intelligent behavior of ants is entirely due to their simple behavior rules, and these rules have the following two characteristics:

1, diversity

2. Positive feedback

Diversity ensures that ants will not go into a dead end when foraging, infinite cycle and positive feedback mechanism ensure that relatively excellent information can be retained. We can regard diversity as a creative ability, while positive feedback is a learning strengthening ability. The power of positive feedback can also be compared to authoritative opinions, and diversity is the creativity to break the authority. It is the careful and ingenious combination of these two points that makes intelligent behavior stand out.

By extension, the evolution of nature, social progress and human innovation are inseparable from these two things. Diversity ensures the innovation ability of the system, and positive feedback ensures the enhancement of excellent characteristics, and the two should be properly combined. If there is too much diversity, that is, the system is too active, which means that ants move too much randomly and will fall into chaos; On the other hand, if the diversity is not enough and the positive feedback mechanism is too strong, then the system is like a stagnant pool. This is manifested in the ant colony's behavior is too rigid, when the environment changes, the ant colony still can not make appropriate adjustments.

Since complexity and intelligent behavior appear according to the underlying rules, and since the underlying rules have the characteristics of diversity and positive feedback, then maybe you will ask where these rules come from? Where does diversity and positive feedback come from? My own opinion: rules come from the evolution of nature. The evolution of nature is also reflected in the ingenious combination of diversity and positive feedback, as I said just now. Why is this a clever combination? Why is the world in front of you so vivid? The answer is that the environment has created all this. The reason why we see a lifelike world is because we can't adapt to the diversity of the environment and the combination of positive feedback has died and been eliminated by the environment! The origin of ant colony algorithm: Ants are one of the most common and abundant insect species on the earth, and often appear in groups in human daily life environment. The swarm intelligence characteristics of these insects have attracted the attention of some scholars. Italian scholars M.Dorigo, V.Maniezzo and others found that ants can always find the shortest path between their nests and food sources when observing their foraging habits. It is found that this group cooperation function of ants is to communicate and coordinate through a volatile chemical called pheromone left in its communication path. Chemical communication is one of the basic ways of information exchange adopted by ants, which plays an important role in their living habits. Through the study of ants' foraging behavior, they found that the whole ant colony cooperated with each other through this pheromone to form positive feedback, which made the ants on multiple paths gradually gather to the shortest path.

In this way, M.Dorigo and others first proposed the ant colony algorithm in 199 1. Its main feature is to find the optimal path through positive feedback and distributed cooperation. This is a heuristic search algorithm based on population optimization. It makes full use of the collective optimization characteristic that biological ant colony can search the shortest path from nest to food through simple information transmission between individuals, and the similarity between this process and solving traveling salesman problem. The optimal solution of NP-hard traveling salesman problem is obtained. At the same time, the algorithm is also used to solve job shop scheduling problem, quadratic assignment problem and multidimensional knapsack problem, which shows its superior characteristics in solving combinatorial optimization problems.

Over the years, researchers all over the world have made serious research and application on ant colony algorithm, which has been widely used in data analysis, robot cooperation problem solving, electric power, communication, water conservancy, mining, chemical industry, construction, transportation and other fields.

Ant colony algorithm can combine the rapidity of problem solving, the characteristics of global optimization and the rationality of the answer in a limited time, so it can attract the attention of researchers in related fields. Among them, the information transmission and accumulation of positive feedback ensure the rapidity of optimization. The premature convergence of the algorithm can be avoided by using its distributed computing characteristics, and the ant colony system with greedy heuristic search characteristics can find an acceptable answer to the question in the early stage of the search process. This superior distributed problem solving model has been greatly improved and expanded on the basis of the original algorithm model through the attention and efforts of researchers in related fields.

After a certain period of time, ants returning from food sources will also encounter obstacles when they reach point D, and they also need to make choices. At this time, the pheromone concentrations on both sides of A and B are the same, or half to the left and half to the right. However, when the ants on the A side have completely bypassed the obstacle and reached point C, the ants on the B side can't reach point C because they have to take a longer path. Figure 3 shows the situation after the ant colony passes in front of the obstacle for a period of time.

At this time, for ants from the nest to point C, because the pheromone concentration on side A is high and the pheromone concentration on side B is low, they tend to choose the path on side A. As a result, there are more and more ants on side A, and finally all ants choose this shorter path. Figure 4 shows the path finally chosen by the ant colony.

The above process is obviously caused by the "positive feedback" process of pheromones left by ants. It is through this information exchange that individual ants achieve the purpose of finding food. The basic idea of ant colony algorithm is also transformed from this process.

Characteristics of ant colony algorithm:

1) ant colony algorithm is a self-organizing algorithm. In system theory, self-organization and other organizations are two basic classifications of organizations. The difference lies in whether organizational forces or organizational instructions come from inside or outside the system, self-organization from inside the system and other organizations from outside the system. If there is no specific external intervention in the process of acquiring space, time or functional structure, the system is said to be self-organizing. In an abstract sense, self-organization is the process of decreasing the entropy of the system without external action (that is, the process of changing the system from disorder to order). Ant colony algorithm fully embodies this process, and takes ant colony optimization as an example to illustrate it. At the beginning of the algorithm, a single artificial ant searches for the solution out of order. After a period of evolution, artificial ants spontaneously tend to find some solutions close to the optimal solution through the role of pheromones, which is a process from disorder to order.

2) Ant colony algorithm is essentially a parallel algorithm. Each ant's search process is independent of each other and only communicates through pheromones. Therefore, agent colony algorithm can be regarded as a distributed multi-agent system, which starts independent solution search at multiple points in the problem space at the same time, which not only increases the reliability of the algorithm, but also makes the algorithm have strong global search ability.

3) Ant colony algorithm is a positive feedback algorithm. It is not difficult to see from the foraging process of real ants that the ants can finally find the shortest path, which directly depends on the accumulation of pheromones on the shortest path, but the accumulation of pheromones is a positive feedback process. For ant colony algorithm, there are exactly the same pheromones in the initial environment, which brings slight disturbance to the system and makes the trajectory concentration of each side different. The solution constructed by ants has advantages and disadvantages. The feedback method adopted by the algorithm is to leave more pheromones on the path of the better solution, and more pheromones attract more ants. This positive feedback process makes the initial difference expand continuously, and at the same time guides the whole system to evolve towards the optimal solution. Therefore, positive feedback is an important feature of ant algorithm, which enables the algorithm to evolve.

4) Ant colony algorithm has strong robustness. Compared with other algorithms, the ant colony algorithm has lower requirements for the initial path, that is, the solution of the ant colony algorithm does not depend on the selection of the initial path, and does not need manual adjustment in the search process. Secondly, the ant colony algorithm has few parameters and simple settings, and is easy to be applied to solve other combinatorial optimization problems.

Application progress of ant colony algorithm Ant colony intelligence represented by ant colony algorithm has become a hot spot in distributed artificial intelligence research, and many algorithms derived from bee colony and ant colony model design have been increasingly applied to the research of enterprise operation mode. The Pentagon is funding the research of swarm intelligence system-swarm strategy. One of its practical uses is to use groups of aerial drones and ground vehicles to divert the enemy's attention and make its troops safely carry out behind the enemy. Based on electronic ants, BT and WorldCom have experimented with new telecommunication network management methods. Swarm intelligence is also applied to the formulation of factory production plan and logistics management of transportation department. Pacific Southwest Airlines adopted a kind of transportation management software directly derived from the research results of ant behavior, and as a result, it saved at least $6,543,800+million annually. Unilever has taken the lead in using swarm intelligence technology to improve the operation of a toothpaste factory. General Motors of the United States, Air Liquide of France, Dutch Ministry of Highway and Transportation and some immigration agencies in the United States also adopt this technology to improve their operational functions. In view of the broad application prospect of swarm intelligence, both the United States and the European Union have started to fund related research projects based on swarm intelligence simulation in recent years, and set up related courses of swarm intelligence in some universities. In China, topics such as evolution, self-adaptation and on-the-spot cognition in the field of swarm intelligence have also been explicitly included in the research content of cognitive science and its information processing, which was given priority support by the National Natural Science Foundation during the Tenth Five-Year Plan period.

Ant colony optimization algorithm was originally used to solve the TSP problem. After years of development, it has gradually penetrated into other fields, such as coloring, large-scale integrated circuit design, routing in communication networks, load balancing, vehicle scheduling and so on. Ant colony algorithm has been successfully applied in many fields, the most successful of which is the application in combinatorial optimization problems.

In the process of network routing, the network traffic distribution is constantly changing, and network links or nodes will randomly fail or rejoin. The self-catalysis and positive feedback mechanism of ant colony are just in line with the characteristics of solving this kind of problem, so ant colony algorithm has been applied in the network field. The parallelism and distribution of ant colony foraging behavior make this algorithm especially suitable for parallel processing. Therefore, the parallel execution of the algorithm has great potential for solving a large number of complex practical application problems.

If there are many unintelligent individuals in a certain group, their intelligent behavior through simple cooperation is called group intelligence. Communication on the Internet is only the result of more neuron connections (human brain) interacting through the Internet, while optical cables and routers are only extensions of axons and synapses. From the perspective of self-organization, there is no essential difference between the intelligence of human brain and that of ant colony. A single neuron has no intelligence, and a single ant has no intelligence, but the connected system is an agent. (Author: Li Jingling Editor: China Electronic Commerce Research Center)