You are on page 1of 7

Hiteshri S.Khandre et al.

/ International Journal of Engineering Science and Technology (IJEST)

REVIEW of APPLICATION of ANT COLONY OPTIMIZATION


Miss. Hiteshri S.Khandre
Computer Tech,RTM Nagpur University,Rajiv Gandhi College Of Engg.,Research & Technology,Babupeth,Chandrapur,India.(+91) hiteshri_khandre@yahoo.com

ABSTRACT Ant colony optimization is a technique for optimization that was introduced in the early 1990's. The inspiring source of ant colony optimization is the foraging behavior of real ant colonies.Ant colony optimization is new meta-heuristic that has proven its quality & versatility on various combinatorial optimization problems such as Traveling Salesman Problem, Vehicle routing problem. The main characteristic of this model are positive feedback, distributed computation, and the use of a constructive greedy heuristic. Positive feedback accounts for rapid discovery of good solutions ,distributed computations avoids premature convergence, and greedy heuristic helps find acceptable solutions in early stages of the search process .In Ant Colony System, a set of cooperating agents called ants cooperate to find good solutions to TSP as well as VRP like problems. Ants cooperate using an indirect communication mediated by a pheromone they deposit on the edges of the graph.This behavior is exploited in artificial ant colonies for the search of approximate solutions to discrete optimization problems, to continuous optimization problems, and to important problems in telecommunications, such as routing and load balancing. First, we deal with the biological inspiration of ant colony optimization algorithms. We show how this biological inspiration can be transferred into an algorithm for discrete optimization. Keywords- Ant Colony Optimization, TSP, VRP, Saving Ants for VRP.

foraging behavior of some ant species. In particular, it was inspired by the double-bridge experiment performed by Jean-Louis Deneubourg and colleagues. In this experiment it was shown that ants are able to find the shortest path to a food source by collectively exploiting pheromones they deposit on the ground while moving. Although ACO has grown to become a fully fledged algorithmic framework and now includes many components that are no longer related to real ants, we report here the doublebridge experiment for its historical value. II.THE DOUBLE-BRIDGE EXPERIMENT In the double bridge experiment, a nest of a colony of Argentine ants is connected to a food source by two bridges. The ants can reach the food source and get back to the nest using any of the two bridges. The goal of the experiment is to observe the resulting behavior of the colony. What is observed is that if the two bridges have the same length, the ants tend to converge towards the use of one of the two bridges. If the experiment is repeated a number of times, it is observed that each of the two bridges is used in about 50% of the cases. These results can be explained by the fact that, while moving, ants deposit pheromone on the ground; and whenever they must choose which path to follow, their choice is biased by pheromone: the higher the pheromone concentration found on a particular path, the higher is the probability to follow that path.

I.INTRODUCTION TO ANT COLONY OPTIMIZATION Ant colony optimization (ACO) is a population-based metaheuristic that can be used to find approximate solutions to difficult optimization problems.In ACO, a set of software agents called artificial ants search for good solutions to a given optimization problem. To apply ACO, the optimization problem is transformed into the problem of finding the best path on a weighted graph. The artificial ants (hereafter ants) incrementally build solutions by moving on the graph. The solution construction process is stochastic and is biased by a pheromone model, that is, a set of parameters associated with graph components (either nodes or edges) whose values are modified at runtime by the ants. The name Ant Colony Optimization was chosen to reflect its original inspiration: the
[1]

Fig.1 - Branches have equal lengths (Deneubourg et al. 1990).

ISSN : 0975-5462

NCICT Special Issue Feb 2011

40

Hiteshri S.Khandre et al. / International Journal of Engineering Science and Technology (IJEST)

Fig.2 - Branches have different lengths (Goss et al 1989).

How Ants Find Food

inserted. Now, two ways are formed to reach the destination (food) .One is longer & another is shorter. Some ants will choose longer path, some will choose shorter one. The ant which is following the shorter path will first to move from nest to food & food to nest. At that time, another ant which is following the longer path will still remains to reach the destination. In this way pheromone concentration will increases on short path & finally most of the ants will prefer that path. The same thing is shown in fig.1 & fig.2. Fig.1 shows symmetric bridges & another one shows asymmetric bridges. The process will be done as explained earlier. III.HEURISTIC To practically solve the large instances one often has to use approximate methods which return near optimal solution in relatively short time .Algorithm of this type is called as Heuristic.
[1]

Obstacle Nest

Food

Fig.3- Figure shows how ants find food.

The ACO metaheuristic is:

This Double-Bridge Experiment is further explained using the above fig.3. A. How ants find food? Ant colony optimization technique is used to solve various combinatorial optimization problems. This technique is based on behavior of ants. The natural tendency of ant is to search for the food. To accomplish this task they work in colony. Some species of ant are blind. So how will they explore the path towards the food? The answer is, Pheromone. Ants naturally release some type of chemical substance which is called pheromone. By smelling that pheromone they will follow each other. Initially, when ants come out from their nest, they randomly will choose any path towards the food. While walking, ants lay down pheromone on that path which they are following. Likewise traveling from source (nest) to food & vice-versa; they laid down pheromone on the path. As shown in above fig.3, we can see ants are moving from source to food. But one obstacle is

Set parameters, initialize pheromone trails SCHEDULE_ACTIVITIES ConstructAntSolutions UpdatePheromones DaemonActions {optional} END_SCHEDULE_ACTIVITIES The metaheuristic consists of an initialization step and of three algorithmic components whose activation is regulated by the SCHEDULE_ACTIVITIES construct. This construct is repeated until a termination criterion is met. Typical criteria are a maximum number of iterations or a maximum CPU time. The SCHEDULE_ACTIVITIES construct does not specify how the three algorithmic components are scheduled and synchronized In most applications of ACO to NP-hard problems however, the three algorithmic components undergo a loop that consists in (i) the construction of solutions by all ants, (ii) the (optional) improvement of these solution via the use of a local search algorithm, and (iii) the

ISSN : 0975-5462

NCICT Special Issue Feb 2011

41

Hiteshri S.Khandre et al. / International Journal of Engineering Science and Technology (IJEST)

update of the pheromones. These three components are now explained in more details. 1) ConstructAntSolutions:Tours are constructed by applying the following procedure to each ant: 1. Choose according to some criterion, a start city at which the ant is positioned. 2. Use pheromone & heuristic values to probabilistically construct a tour by iteratively adding the cities that ants have not visited yet, until all cities have been visited. 3. Go back to initial city. Each ant located at city i hops to a city j selected among the cities that have not yet been visited according to the probability, (Dorigo et al. 1991, 1996): Pk(i,j) = ij *ij/ gJk(i,j) ig *ig Pk(i,j): Probability that ant k in city i will go to city j gJk (i, j): Set of cities that have not yet been visited by ant k in city i Where ij and ij are respectively the pheromone value and the heuristic value associated with the . Furthermore, and are component positive real parameters whose values determine the relative importance of pheromone versus heuristic information. 2) DaemonActions: Once solutions have been constructed, and before updating the pheromone values, often some problem specific actions may be required. These are often called daemon actions, and can be used to implement problem specific and/or centralized actions, which cannot be performed by single ants. The most used daemon action consists in the application of local search to the constructed solutions: the locally optimized solutions are then used to decide which pheromone values to update. 3) UpdatePheromones: The aim of the pheromone update is to increase the pheromone values associated with good solutions, and to decrease those that are associated with bad ones.

Usually, this is achieved (i) by decreasing all the pheromone values through pheromone evaporation, and (ii) by increasing the pheromone levels associated with a chosen set of good solutions Pheromone evaporation implements a useful form of forgetting, means after some pheromone will get evaporated, favoring the exploration of new areas in the search space. Different ACO algorithms, for example ant colony system (ACS) (Dorigo & Gambardella 1997) or MAXMIN ant system (MMAS) (Sttzle & Hoos 2000), differ in the way they update the pheromone. IV.MAPPING OF TSP & ACO ALGORITHM In all available ACO algorithms for TSP, the pheromone trails are associated with the arcs & therefore ij refers to the desirability of visiting city j directly after city i. The heuristic information is chosen as ij=1/dij i.e., heuristic desirability of going from city i to city j is inversely proportional to the distance between the cities.1) dij = 0 ,the set of ij is set to very small value. For implementation purpose, pheromone trails are collected into a pheromone matrix whose elements are the ij. This can be analogous to heuristic information. V.[3] VEHICLE ROUTING PROBLEMS A. Introduction The European situation in freight transportation reflects the need for improved efficiency, as the traffic volume increases much faster than the street network grows. Thus, given the current efficiency, this will eventually lead to a breakdown of the system. However, with rapidly increasing computational power intelligent optimization methods can be developed and used to increase the efficiency in freight transportation and circumvent the above mentioned problem.The VRP involves the design of a set of minimum cost delivery routes, originating and terminating at a depot, which services a set of customers. Each customer must be supplied exactly once by one vehicle route. The total demand of any route must not exceed the vehicle capacity. The total length of any route must not

ISSN : 0975-5462

NCICT Special Issue Feb 2011

42

Hiteshri S.Khandre et al. / International Journal of Engineering Science and Technology (IJEST)

exceed a pre-specified bound. This problem is known to be NP-hard,such that exact methods like Dynamic Programming or Branch & Bound work only for relatively small problems in reasonable time.Thus, a large number of approximation methods have been proposed. Most of the recent approaches are based on meta-heuristics like Tabu Search, Simulated Annealing and Ant Systems. The Ant System approach, belonging to a class of methods called Ant Colony Optimization, which is mentioned above. This approach has been applied to a number of combinatorial optimization problems, such as the Graph Coloring Problem , the Quadratic Assignment Problem, the Travelling Salesman Problem , the Vehicle Routing Problem ([4], [5]) and the Vehicle Routing Problem with Time Windows ([6]). Recently, a convergence proof for a generalized Ant System has been developed by Gutjahr ([7]).In the previous approaches for the VRP ([4], [5]) the construction of solutions was based on a sequential tour building approach, which utilized a parametrized savings criterion. The main idea of the new approach is to transfer the simultaneous tour construction mechanism proposed in [8] into a rank based Ant System.Our computational findings show that a considerable improvement is achieved through this new approach. The Savings algorithm is described by author. VI.SAVINGS AND ANT SYSTEM ALGORITHMS FOR VRPS The basic structure of Ant System algorithm is identical to the one proposed in [5]. A. The Savings algorithm, proposed in [8], is the basis of most commercial software tools for solving VRPs in industrial applications. It is initialized with the assignment of each customer to a separate tour. After that for each pair of customers i and j the following savings measure is calculated: s (i, j) = d(i,0) + d(0, j) - d(i, j) where d(i, j) : the distance between locations i and j and the index 0 denotes the depot. Thus, the values s(i, j) : the savings of combining two

customers i and j on one tour as opposed to serving them on two different tours. In the iterative phase, customers or partial tours are combined according to these savings, starting with the largest savings, until no more combinations are feasible. A combination is infeasible if it violates either the capacity or the tourlength constraints. The result of this algorithm is a (sub-) optimal set of tours through all customers. B. The Ant System algorithm for the VRP Bullnheimer et al. ([4], [5]) have first applied the Ant System to the VRP. Their approach centers on the similarity of VRPs with TSPs, namely the fact that for a given clustering of customers the problem reduces to several TSPs. Thus, their Ant System is strongly influenced by the Ant System algorithms applied to the TSP. On the contrary, approach, to the best knowledge, is the first combination of a heuristic algorithm for the VRP with an Ant System. The Ant System algorithm mainly consists of the iteration of three steps: 1) Generation of solutions by ants according to private and pheromone information 2) Application of a local search to the ants solutions 3) Update of the pheromone information In addition to that new approach features a fourth step, namely: 4) Augmentation of the Attractiveness list, which stores the desirability of all feasible combinations. The implementation of these four steps is described below. C. Generation of solutions As stated above, the solution generation technique which was implemented by author is the main contribution of this paper. So far, in Ant Systems solutions for the VRP have been built using a Nearest Neighbor heuristic (see e.g. [4], [5]). As opposed to that we use the Savings algorithm described above to generate solutions. For that ,some modifications has been done in Ant System.This modification is done in the following way. First, a sorted list of attractiveness values is generated in decreasing order.

ISSN : 0975-5462

NCICT Special Issue Feb 2011

43

Hiteshri S.Khandre et al. / International Journal of Engineering Science and Technology (IJEST)

Initially, we generate a sorted list of attractiveness values in decreasing order. These attractiveness values feature both the savings values as well as the pheromone information. Thus the list consists of the following values: ij = [s(i,j)] [ij] where ij denotes the pheromone concentration on the arc connecting customers i and j, The pheromone concentration ij contains information about how good the combination of two customers i and j was in previous iterations. In each decision step of an ant, we consider the k best combinations still available, where k is a parameter of the algorithm which will be refered as neighborhood below. Let k denote the set of k neighbors, i.e. the k feasible combinations (i,j) yielding the largest savings, considered in a given decision step, then the decision rule is given by equation Given below, where Pij is the probability of choosing to combine customers i and j on one tour. Pij = ij / (h,l) k hl The construction process is stopped when no more feasible combinations are possible. 1) Local Search: Following [5], the 2-opt algorithm (c.f. [9]) is applied to all vehicle routes built by the ants, before the pheromone information is updated. The 2-opt algorithm was developed for the traveling salesman problem and iteratively exchanges two edges with 2 new edges until no further improvements are possible. 2) Pheromone update:After all ants have constructed their solutions, the pheromone trails are updated on the basis of the solutions found by the ants. According to the rank based scheme proposed in the pheromone update is given by two procedures: 1) Pheromone deposition 2) Pheromone evaporation ij new= ijold + k=1m ijk + ebsij where 01 is the trail persistance and e is the number of elitists. Using this scheme two kinds of trails are laid. First, the best solution found during the process is updated as if elitist ants had

traversed it. The amount of pheromone laid by the elitists is ijbs = 1/cbs where cbs is the objective value of the best solution found so far. Second, the m best ants of the iteration are allowed to lay pheromone on the arcs they traversed. The quantity laid by these ants depends on their rank k as well as their solution quality ck such that the k-th best ant lays ijk = (m-k) / ck .Arcs belonging to neither of those solutions just lose pheromone at the rate (1-), which constitutes the trail evaporation. D. Augmentation of the Savings list After the pheromone information has been updated the attractiveness values ij are augmented with the new pheromone information. After the augmentation the attractiveness values are again sorted in decreasing order. This mechanism is the second important contribution of new approach. In the beginning the attractiveness values are sorted according to the savings values, as the pheromone is equal on all arcs. As learning occurs, and some arcs are reinforced through the update of the pheromone information, the attractiveness values ij change, as they become more and more biased by the pheromone information. Thus, values that was initially high but turned out not to be in good solutions will decrease, while combinations with initially low values that appeared in good solutions will become more attractive. As the attractiveness values are re-sorted after each iteration, this leads to dynamic effects. In particular, good arcs are reinforced twice. First, they receive more pheromone than others, and second as their attractiveness increases they are considered earlier in the constructive process. E. Parameter settings In order to keep the results comparable and to isolate the effects of new approach, basically, they have used the same parameter values. Thus, n artificial ants are used, ==5 and m = 6 elitist ants. For this approach an evaporation rate = 0.95 is preferable to = 0.75 as proposed in earlier works. Apart from that we varied the number of iterations in order to be able to estimate the performance of algorithm for different run times.

ISSN : 0975-5462

NCICT Special Issue Feb 2011

44

Hiteshri S.Khandre et al. / International Journal of Engineering Science and Technology (IJEST)

More specifically, results for n/2,n,2*n iterations together with the corresponding computation times is provided. Apart from that, given the dynamic savings list, the run with different sizes of the neighborhood is performed, i.e. with different numbers of alternatives in each decision step of an ant. Thus, results for k=n/5, k=n/4, k=n/2, k=n is provided in this paper. F. Comparison between new approach and existing meta-heuristics Now that we know the appropriate size of the neighborhood k =n/4, we will compare the best results obtained with new approach (denoted by SavingsAnts) after 2.n iterations using this neighborhood size with the results of the previous Ant System algorithm for the VRP as well as with Tabu Search and Simulated Annealing algorithms. VII.CONCLUSION In this paper, at the beginning, the basic concept of Ant Colony Optimization is explained with Double-Bridge Experiment. Then, how ACO metaheuristic algorithm works, is explained. Mapping of ACO with Traveling Salesman Problem is given.In further segment, the possible improvements to standard Ant System approaches for VRPs through the use of a problem specific heuristic, namely the Savings algorithm is given.In which the computational study performed shows the superior performance of new approach over the existing Ant System algorithm (Like Tabu Search and Simulated Annealing). In particular, the average result of new approach improves the solution quality of the previous Ant System significantly. REFERENCES
[1]www.scholarpedia.org/article/Ant_colony_optimization [2] staff.washington.edu [3] Costa, D. and Hertz, A.: Ants can colour graphs. Journal of the Operational Research Society 48(3) (1997) 295305 [4] Bullnheimer, B., Hartl, R. F. and Strauss, Ch.: Applying the ant system to the vehicle routing problem. In: Voss, S., Martello, S., Osman, I. H. and Roucairol, C. (Eds.):MetaHeuristics: Advances and Trends in Local Search Paradigms for Optimization. Kluwer, Boston (1999) [5] Bullnheimer, B., Hartl, R. F. and Strauss, Ch.: An improved ant system algorithm for the vehicle routing problem. Annals of Operations Research 89 (1999) 319328

[6] Gambardella, L. M., Taillard, E. and Agazzi, G.: MACSVRPTW: A Multiple Ant Colony System for Vehicle Routing Problems with TimeWindows. In: Corne, D., Dorigo, M. and Glover, F. (Eds.): New Ideas in Optimization. McGraw-Hill, London (1999) [7] Gutjahr, W. J.: A graph-based Ant System and its convergence. Future Generation Computing Systems. 16 (2000) 873888 [8] Clarke, G. and Wright, J. W.: Scheduling of vehicles from a central depot to a number of delivery points. Operations Research 12 (1964) 568581 [9] Croes, G. A.: A method for solving Traveling Salesman Problems. Operations Research 6 (1958) 791801 Science 40 (1994) 12761290

ISSN : 0975-5462

NCICT Special Issue Feb 2011

45

Copyright of International Journal of Engineering Science & Technology is the property of Engg Journals Publications and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.