See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/220792786

An Ant Colony Optimization Algorithm for Multiple Travelling Salesman

Problem.

Conference Paper · January 2006

DOI: 10.1109/ICICIC.2006.40 · Source: DBLP

CITATIONS READS
65 2,011

2 authors, including:

Dingwei Wang
Northeastern University (Shenyang, China)
141 PUBLICATIONS   1,174 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Dingwei Wang on 30 July 2015.

The user has requested enhancement of the downloaded file.

 An Ant Colony Optimization Algorithm for
Multiple Travelling Salesman Problem

Pan Junjie1 and Wang Dingwei2

1
School of Information Science and Engineering,
Northeastern University,Shenyang,110004, China
[email protected]
2
School of Information Science and Engineering,
Northeastern University ,Shenyang ,110004, China
[email protected]

Abstract this problem. Good heuristic techniques are necessary

for solving MTSP due to its high computational
Multiple travelling salesman problem (MTSP) is a complexity. Modern heuristic techniques, namely
typical computationally complex combinatorial genetic algorithms and ant colony optimization,can be
optimization problem,which is an extension of the good candidates for this problem.
famous travelling salesman problem (TSP). The MTSP In this research, ant colony optimization (ACO)
can be generalized to a wide variety of routing and heuristic is used for solving ability limited MTSP. The
scheduling problems.It is known that classical main research contribution of the present paper is to
optimization procedures are not adequate for this make the first attempt in the published literature to
problem.The paper makes the attempt to show how the show how the ACO algorithm can be applied to MTSP
ant colony optimization(ACO) can be applied to the with the limited ability,and the results are encouraging.
MTSP with ability constraint. In this paper,we compare In the following sections of this paper, the
it with MGA by testing several standard problems algorithm based ACO for MTSP is explained then the
from TSPLIB.The computational results show that the computational results are reported.
proposed algorithm can find competitive solutions
even not all of the best solutions within rational time, 2. The MTSP with ability constraint
especially for large scale problems.
The MTSP can be stated as follows:There are m
1. Introduction salesman who must visit a set of n cities,and each
salesman is defined to start and end at the same
Travelling Salesman Problem(TSP) is the most depot.In this problem,each city must be visited exactly
famous and well-studied problem in the combinatorial once by only one salesman and its objective is to find
optimization area[1].And the mutiple travelling the minimum of total distances travelled by all the
problem(MTSP) is an extension of TSP.This problem salesmen.An example is depicted in Fig.1,where
relates to accommodating real world problems where m=3,n=7.Several authors[6,7] suggested transforming
there is a need to account for more than one the MTSP with m salesmen and n cities into a TSP
saleman.The MTSP can be generalized to a wide with n+m-1 cities by the introduction of m-1 artificial
variety of routing and scheduling problems,for depots(n+1,…,n+m-1).The tranformation of the
example,the School Bus Routing Problem[2,3]and the previous example is depicted in Fig.2.
Pickup and Delivery Problem[4,5].Therefore,finding a Several authors have some researchs on
good optimal solution method for the MTSP is MTSP.Samerkae Somhom et.al. have used
important and induces to improve the solution of any Competition-based neural network to solve MTSP with
other complex routing problems.However,MTSP is a minmax objective[8],and LinxinTang et.al. have used
NP-complete problem for which optimal solutions can the modified genetic algorithm to solve hot rolling
only be found for small size problems. It is known that scheduling problem,which is an example of MTSP[9].
classical optimization procedures are not adequate for

Proceedings of the First International Conference on Innovative Computing, Information and Control (ICICIC'06)
0-7695-2616-0/06 $20.00 © 2006
 However,the resulting TSP is highly Based the general MTSP,the target function can be
degenerate,when an MTSP is transformed to a single described as follows:
TSP since the resulting problem is more arduous to m
solve than an ordinary TSP with the same number of
cities. While the general objective of the MTSP is to
Min ¦ td
i =1
i (2)
minimize the total distance which can be called In this way,we can minimize the total distance on
minimum criterion,generally,there are m-1 cities the premise of the limited ability of salesmen.
always to select the nearest cities as their round trip.As
a result,TSP which is made up of the left n-m+1cities is
left.During the m salesmen,there are m-1 salesmen
3. Ant Colony Optimization
travelling only one city,and one saleman needs to
Ant colony algorithms are becoming popular
travel the left n+m-1 cities. This is not up to the
mustard.In practice,every saleman has the similar
approaches for solving combinatorial optimization
ability,and the limit in ability.So the MTSP with ability
problems in the literature.They were first introduced by
constraint is more appropriate in the real world
Dorigo et al[11].The fundamental idea of ant heuristics
problems.In this paper,we suppose the number of cities
is based on the behabiour of natural ants that succeed
in finding the shortest paths from their nest to food
which are travelled by every salesman is limited.
sources by communicating via a collective memory
5 7 that consists of pheromone trails.Due to ant’s weak
global perception of its environment,an ant moves
essentially at random when no pheromone is
3 available.However,it tents to follow a path with a high
4
1 pheromone level when many ants move in a common
area,which leads to an autocatalytic process.Finally,the
ant does not choose its direction based on the level of
pheromone exclusively,but also takes the proximity of
2 the nest and of the food source,respectively,into
6 account.This allows the discovery of new and
potentially shorter paths.
Fig 1. Example solution of MTSP
4. ACO for the MTSP with ability
1
constraint
7 5
4.1 Solution construction

3 4 Some authors used ACO to solute TSP,however,

MTSP is different from TSP.In thepaper,we improve
on ACO according to the characteristic of the
8 MTSP.One salesman should firstly travel an amount of
9 cities,and the next salesman travels an amount of
6 2 unvisited cities,in this way,all the salemen travell the
total cities. The travelled number of salesman is
Fig 2. Transformation from MTSP to TSP generated randomly in a certain range.Define the
number of cities salesman i travell is tni ,then
We define the following variables and parameters:
­2 ≤ tni ≤ l i
Distance from city i to city j: °m
Dij, i ∈ {1,...n + m − 1},j ∈ {1,...,n + m − 1} ® (i = 1,2,..., m) (3)
°¦ i tn = n − 1
li is maximal number of cities salesman ¯ i
i(1 ≤ i ≤ M ) can travel,ni is the real number of cities the Where m is the number of salesmen,n is the numer
salesman i(1 ≤ i ≤ M ) has travelled, of cities, l i is the max number of cities saleman i can
tdi is the total distance salesman i(1 ≤ i ≤ M ) has travel.
travelled,define the penalty function is In the ACO algorithm,we assign a tour list to every
if ni>li td i = ( ni − li + 1) * td i (1) ant.Evey ant tours from the start city,and then select

Proceedings of the First International Conference on Innovative Computing, Information and Control (ICICIC'06)
0-7695-2616-0/06 $20.00 © 2006
 unvisited cities which is not start city or artificial IJ ij (t + n) = ȡIJ ij (t) + ǻIJ ij (5)
cities.We suppose that the saleman i is visiting when
ants start from the start city or artificial cities the ith where ȡ is a parameter that controls the speed of
time.While the salesman i is visiting,if the number of evaporation, ǻIJ ij is the adding pheromone to all the
cities the ants have visited is equal to tni,then the ants
arcs in this tour.
select a unvisited artificial city,it means that the next M
salesman starts to visit.The cities which have been ΔIJ ij = ¦ ΔIJ ijk (6)
visited are recorded in the tour list ,until all the cities k =1
are visited by all the salesmen.In this way,a valid
Where ΔIJ
k
solution is constructed.In the algorithm,we used k ij is the adding pheromone to the arcs in the
ants,so there are k solutions constructed. tour ant k has visited. if i,j are at the tour ant k has

4.2 Route selection visited.

ΔIJ ijk = Q , (7)
Every ant selects the next city independently. Lk
Suppose that the probability of ant k moves from city i
k Where Q is a constant, Lk is the distance ant k has
to city j is Pij (t) ,uses the following probabilistic
toured.
formula: This updating encourages the use of shorter routes
and increases the probability that future routes will use
­ IJ ij (t) Į Ș ijȕ the arcs contained in the best solutions. This process is
, j∈ A
°°
[ ]
Pijk (t) = ® ¦ IJ ij Ș il
Į ȕ
(4) repeated for a predetermined number of iterations and
° l ∈ Ak the best solution from all of the iterations is presented
°¯ 0 , otherwise as an output of the model and should represent a good
approximation of the optimal solution for the problem.
Where Șij is an optional weighting function, The
5. Computational experiments
values that are given by the weighting function are
commonly called the heuristic information,here, To demonstrate the efectiveness of the algorithm
−1
suppose Șij = ( d ij ) , d ij is the distance from city i to computational experiment is conducted on a series of
standard data. The data set which is selected from
city j. Ak = N\tabu k (t) , Ak denotes the cities that ant k theTSPLIB ,and we suppose the number of salesmen is
has not visited,and tabu k (t) denotes that cities that ant k 5. The algorithms are coded by C++ language and
has visited. Furthermore, the exponents α and β are implemented on a Pentium 4 PC at 2.66GHZ(512MB
RAM).
positive parameters whose values determine the In order to evaluate the efficacy of the ACO
relation between pheromone information and heuristic algorithm for the MTSP with ability constraint, we
information. simulated the other algorithm available in the
literature, namely modified genetic
4.3 Trail updating algorithm(MGA)[9]. The algorithms are applied to 6
standard problems, for most of which the optimal
In order to improve future solutions, the pheromone solutions are available. The results of this simulation
trails of the ants must be updated to reflect the ant’s are given in Table 1. In this table, for each algorithm
performance and the quality of the solutions the average and the best of obtained solutions through
found.This updating is a key element to the adaptive 10 replications as well as the computation time is
learning technique of ACO and helps to ensure reported.The table shows that MGA is little superior to
improvement of subsequent solutions. Trail updating the proposed ACO algorithm in pr76 and pr152,which
includes local updating of trails after individual are not large scale problems.And in following larger
solutions have been generated and global updating of scale problems,such as pr226,pr299,pr439 and
the best solution route after a predetermined number of pr1002,the solutions of the proposed ACO algorithm
solutions m has been accomplished. This is done with are superior to MGA.
the following local trail updating equation,

Proceedings of the First International Conference on Innovative Computing, Information and Control (ICICIC'06)
0-7695-2616-0/06 $20.00 © 2006
 Table 1. The computational results

[3] Orloff DS,Routing a fleet of M vehicles to/from a central

6. Conclusion facility,Networks,1974,4.147-162.

[4] Christofides N,Eilon S,An algorithm for the vehicle

ACO is a promising optimization technique for dispatching problem,Operations Research Quarterly,1969,
solving complex combinatorial optimization problems 20.309-318.
like the MTSP. In the past, ACO has been applied to
several combinatorial optimization problems [5] Savelsbergh MWP,The general pickup and delivery
successfully. In this study its application to the MTSP problem,Transactions of Science ,1995,29.17-29.
with ability constraint which is an extension of the
well-known TSP is presented. To evaluate its [6] Sveska JA,Huckfeldt VE,Computational experience with
effectiveness,we compare the algorithm with MGA by an m-salesman traveling salesman algorithm.Management
Science,1973,19.790-799.
testing problems from standard TSPLIB. In these tests
the proposed algorithm found not all of the best [7] Bellmore M,Hong S,Transformation of multisalesmen
solutions but competitive solutions within rational problem to the standard traveling salesman problem,Jurnal of
time,especially in large scale problems. the Association Computer Machinery,1974,21.500-504.
Ant colony optimization clearly has the ability to
find good results within 1% of the known optimum for [8] Samerkae Somhom,Competition-based neural network
small problems. However, consistent with past for the multiple travelling salesmen problem with minmax
research, the ACO methods used in this research are objiective,Computers&Operations,1999,395-407.
not as efficient in finding solutions for larger
[9] Lixin Tang,Jiyin Liu,multiple traveling salesman problem
problems.In the research,a algorithm based ACO is model for hot rolling scheduling in Shanghai Baoshan Iron &
applied to larger problems successly,and the solution is Steel Complex,European Journal of Operational Research ,
encouraging, 2000,24.267- 282.

7. References [10]Lenstra JK,Rinnooy Kan AHG,Some simple applications

of the travelling salesma n problem,Operations Research
[1] Lawler EL,Lenstra JK,Kan Rinnooy AHG,Shmoys DB, Quarterly,1975,26.717-734.
The traveling salesman problem:a guided tour of
combinatorial optimization,Chichester,Wiley,1985, 11.201- [11] Dorigo M,Maniezzo V,Colomi A,Positive feedback as a
209. search strategy,Technical Report,Politecnico idi Milano,
1991,91-106.
[2] Angel RD,Caudle WL,Noonan R,Whinson A,Computer
assisted school bus scheduling,Management Science,1972,
18.279-88.

Proceedings of the First International Conference on Innovative Computing, Information and Control (ICICIC'06)
0-7695-2616-0/06 $20.00 © 2006

View publication stats