PLOS ONE

RESEARCH ARTICLE

An improved ant colony optimization

algorithm based on context for tourism route
planning
Shengbin Liang ID1, Tongtong Jiao ID1, Wencai Du2, Shenming Qu1*

1 School of Software, Henan University, Kaifeng, Henan, China, 2 Institute for Data Engineering and Science,
University of Saint Joseph, Macau SAR, China

a1111111111 * [email protected]
a1111111111
a1111111111
a1111111111 Abstract
a1111111111
To solve the problem of one-sided pursuit of the shortest distance but ignoring the tourist
experience in the process of tourism route planning, an improved ant colony optimization
algorithm is proposed for tourism route planning. Contextual information of scenic spots sig-
OPEN ACCESS nificantly effect people’s choice of tourism destination, so the pheromone update strategy is
Citation: Liang S, Jiao T, Du W, Qu S (2021) An combined with the contextual information such as weather and comfort degree of the scenic
improved ant colony optimization algorithm based spot in the process of searching the global optimal route, so that the pheromone update
on context for tourism route planning. PLoS ONE tends to the path suitable for tourists. At the same time, in order to avoid falling into local
16(9): e0257317. https://doi.org/10.1371/journal.
optimization, the sub-path support degree is introduced. The experimental results show that
pone.0257317
the optimized tourism route has greatly improved the tourist experience, the route distance
Editor: Diego Oliva, Universidad de Guadalajara,
is shortened by 20.5% and the convergence speed is increased by 21.2% compared with
MEXICO
the basic algorithm, which proves that the improved algorithm is notably effective.
Received: July 4, 2021

Accepted: August 30, 2021

Published: September 16, 2021

Peer Review History: PLOS recognizes the

benefits of transparency in the peer review
Introduction
process; therefore, we enable the publication of With the development of mobile Internet technology, the user travel information is becoming
all of the content of peer review and author more and more diversified. Travel notes, strategies, short videos and so on have become the
responses alongside final, published articles. The
main basis for the user tourism route planning.
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0257317 While tourism route planning is a kind of loop problem that tourists start from the starting
point, pass through the selected scenic spots only once, and finally return to the starting point.
Copyright: © 2021 Liang et al. This is an open
Similar to (Travelling Salesman Problem)TSP, tourism route planning is a NP-Hard problem,
access article distributed under the terms of the
Creative Commons Attribution License, which but it is more complex and affected by many constraints. When tourists arrive in a city to visit
permits unrestricted use, distribution, and the scenic spots in its territory, they are often affected by time, transportation conditions, eco-
reproduction in any medium, provided the original nomic cost and other factors. In practice, people often spend a lot of time on the selection of
author and source are credited. scenic spots and route planning.
Data Availability Statement: All relevant data are ACO algorithm is proposed by Dorigo et al. [1] according to the intelligent behavior of ant
within the manuscript. colony in the process of foraging. The algorithm has some advantages such as heuristic, posi-
Funding: Author: Liang Shengbin Grant number:
tive feedback and distributed. ACO algorithm has been widely used to solve (Traveling Sales-
No.20A520008 Funder: Key Scientific Research man Problem)TSP. Stutzle et al. [2] proposed Max-Min ant system (MMAS) to solve the
Projects of Colleges and Universities in Henan problem but the improved algrithm is easy to fall into local optimum; Yang et al. [3] proposed

PLOS ONE | https://doi.org/10.1371/journal.pone.0257317 September 16, 2021 1 / 16

PLOS ONE An improved ant colony optimization algorithm based on context for tourism route planning

Province URL: http://jyt.henan.gov.cn/2019/07-31/ an improved ACO algorithm based on game theory, which introduced entropy weighted
1604951.html Did the sponsors or funders play learning strategy to optimize the accuracy of the optimal solution of (Traveling Salesman Prob-
any role in the study design, data collection and
lem)TSP problem on the basis of ACO and MMAS. Deng et al. [4] divided the optimization
analysis, decision to publish, or preparation of the
manuscript? No. problem into several sub problems in order to improve the convergence rate of ACO algo-
rithm and the pheromone update strategy was used to improve the optimization ability, then
Competing interests: The authors declare that no
coevolution mechanism was used to exchange information among different sub populations,
competing interests exist.
so as to avoid the ant colony falling into local optimum. Qian et al. [5] took the traffic cost as
the calculation object, optimized the automatic adjustment mechanism of ant colony phero-
mone, and improved the performance of ACO algorithm. In addition, some researchers com-
bined ACO algorithm with other algorithms. For example, Liang et al. [6] combined genetic
algorithm with ACO algorithm to further improve the ability to solve the defects of local opti-
mization, obtain the best path and save cost. Che et al. [7] combined particle swarm optimiza-
tion with ACO algorithm to find the optimal path and improved the quality of path planning
by improving pheromone update rules and heuristic function based on particle swarm
optimization.
Scholars have put forward a large number of improvement methods and strategies to solve
various problems in different boundaries, different disciplines and different fields. Aiming at
the problems of local optimization and slow convergence of ACO algorithm, Wang et al. [8]
embedded genetic algorithm and cloud model into ACO algorithm to obtain the optimal solu-
tion. Han et al. [9] proposed an an algorithm called SOS-ACO that combines symbiotic organ-
isms search (SOS) and ACO algorighm to calculate the optimal or near-optimal assembly
sequence, which find the optimal or near optimal assembly sequence in fewer iterations. ACO
algorithm is also widely used in feature selection. Zhou et al. [10] proposed a two-stage hybrid
ACO for high-dimensional feature se-lection (TSHFS-ACO), which uses the interval strategy
to determine the size of OFS for the following OFS search and helps to reduce the complexity
of the algorithm and alleviate the algorithm from getting into a local optimum. In solving con-
straint satisfaction problem (CSP), in order to overcome the shortcomings of low solution
quality and slow convergence speed based on ACO algorithm, Guan et al. [11] proposed an
improved ant colony optimization with an automatic updating mechanism (AU-ACO). In
order to take advantage of both epsilon greedy and Levy flight, Liu et al. [12] proposed a
greedy-Levy ACO incorporating these two approaches to solve complicated combinatorial
optimization problems, which outperforms max-min ACO and other latest solvers. It is useful
to implement collective intelligence(CI) evolution using ant colony optimization (ACO), Gan
et al. [13] analyzed the performance of ACO evolution algorithm and verified the feasibility of
applying the collective intelligence (CI) evolution theory to a specific application.
Besides ACO algorithm, traditional route planning methods include tabu search algorithm
[14], genetic algorithm [15, 16], particle swarm optimization algorithm [17], simulated anneal-
ing algorithm [18] etc. ACO algorithm is a parallel algorithm, the search process of each ant is
independent, and ants communicate through pheromone. Therefore, ACO algorithm can be
regarded as a distributed multi-agent system, which starts to search for independent solutions
at multiple points in the problem space at the same time, which not only increases the reliabil-
ity of the algorithm, but also makes the algorithm have strong global search ability. In this
paper, a solution method of tourism route planning problem based on context feedback and
ACO algorithm is proposed.
In the above studies, the shortest distance is used as the goal of route planning. However, in
the tourism route planning, in addition to the main factors that affect the route planning, the
waiting time (comfort degree of the scenic spot) and the weather also have an important
impact on the route planning. In ACO algorithm, ants communicate with each other through
pheromones, and use learning mechanism to adjust the selection probability of the optimal

PLOS ONE | https://doi.org/10.1371/journal.pone.0257317 September 16, 2021 2 / 16

PLOS ONE An improved ant colony optimization algorithm based on context for tourism route planning

path. However, the ACO algorithm also has some defects, such as the lack of pheromone in
the initial stage of path construction, the evolution speed is very slow; with the ACO algorithm
using positive feedback principle to strengthen the optimal solution, after a certain number of
iterations, pheromone is mainly concentrated on a few routes, resulting in premature conver-
gence. At the same time, from the perspective of adjusting the load of scenic spots, the routes
to individual scenic spots also lead to a sharp increase in the number of tourists, resulting in a
sharp decline in tourist experience and a longer waiting time.
We propose a tourism route planning method based on contextual information feedback
mechanism and ant colony optimization algorithm in this paper. The major works involved in
this paper are as follows:
1. Take distance, scenic spots comfort ratio, as comprehensive evaluation indicators to
improve tourists experience while ensuring economic costs.
2. The tourism route should pass through all the scenic spots selected by tourists, only once,
and finally return to the starting point.
3. Introduce the comfort degree of the scenic spot, weather and other contextual information
into the improved algorithm, and dynamically adjusts the tourism route planning through
self-learning.
4. From the perspective of tourists, the tourism route planning should be people-oriented,
reduce the economic cost of tourists and time, enhance the travel experience of tourist.
This paper proceeds in the following order. In section II, introduce ACO algorithm and its
lastest researches. In section III, the proposed ACOCA algorithm is described in detail, includ-
ing contextual information modelling, and ACOCA algorithm description. In section IV, a
series of performance tests are simulated and the results of the experiments are analyzed. In
section V, the conclusions and our further research directions are given.

ACO algorithm
ACO algorithm is a kind of swarm intelligence algorithm. By simulating the process of ant col-
ony foraging, the shortest path in different environments is established by using the internal
information transmission mechanism of ant colony. In the process of activity, ants will leave
pheromone, and the subsequent ants can choose the path according to the pheromone left by
the previous ants. The higher the concentration of pheromone remaining on the path, the
higher the probability that the ant will choose this path. At the same time, the concentration of
pheromone will volatilize with time. Therefore, through the ant colony behavior, ants continu-
ously learn and optimize through the information feedback mechanism to determine the
shortest foraging path. According to the characteristics of ACO algorithm, it has been widely
used in path planning [19], network routing [20], logistics distribution [21], trip route plan-
ning [22] and traveling salesman problem [6, 23].
For the studies of the ACO algorithm, some researchers proposed a lot of improved meth-
ods and applications in recent years. Yang et al. [24] proposed an improved ant colony optimi-
zation algorithm to control an output scattered light field after an incident beam of light passes
through a turbid medium. Yu et al. [25] proposed an ant colony algorithm based on magnetic
neighborhood and filtering recommendation (MRACS) to deal with the problems that the
ACO algorithm has slow convergence speed and easy falling into the local optimum when
solving (Traveling Salesman Problem)TSP. Wu [26] proposed a hybrid ant colony optimiza-
tion (HACO) algorithm for solving the problem of vehicle routing with time windows. In view
of the shortcomings of traditional ACO algorithm in path planning of indoor mobile robot,

PLOS ONE | https://doi.org/10.1371/journal.pone.0257317 September 16, 2021 3 / 16

PLOS ONE An improved ant colony optimization algorithm based on context for tourism route planning

such as a long time path planning, non-optimal path for the slow convergence speed, and local
optimal solution characteristic of ACO, Miao et al. [27] proposed an improvement adaptive
ant colony optimization (IAACO) algorithm to address these problems. Jia et al. [28] proposed
an ant colony-based algorithm for integrated scheduling on batch machines with non-identical
capacities. Cerda et al. [29] proposed an algorithm based on ACO algorithm metaheuristics to
dynamically optimize the decision thresholds provided by the Pairs Trading investment strat-
egy. The control method of space vector pulse width modulation (SVPWM) overmodulation
region II has the disadvantages of a complicated process and large harmonic content. To solve
these problems, Zhang [30] proposed an SVPWM overmodulation region II control method
based on the chaos ant colony algorithm. To address the lack of convergence speed and diver-
sity of ACO algorithm, Yu et al. [31]proposed a dynamic reproductive ant colony algorithm
based on piecewise clustering (RCACS) to optimize the problems. Xiao et al. [32] proposed an
improved ACO algorithm with a negative feedback mechanism to adress the ACA encounters
difficulty escaping from the local optimal solution. Based on ant colony optimization algo-
rithm, Wang et al. [33]constructs a sports competition venue simulation method based on ant
colony optimization algorithm and artificial intelligence to improve the reduction degree of
sports competition venue simulation. Kanso et al. [34] proposed an Hybridized Ant Colony
Algorithm (HACA) and hybridized with a local search algorithm that involves the 2-opt,
Swap, Relocate and Cross-exchange moves to solve OCARP problem. Shi [35] proposed a
routing protocol and simulated based on an ant colony optimization algorithm’s performance
for reducing the energy consumption of sensors in IoT networks for increasing network
lifespan.
As a meta-heuristic algorithm, ACO algorithm is a candidate method in supply chain man-
agement (SCM). According to Axsater [36, 37] and Mostafa et al. [38], the cost functions of
two or three-echelon inventory systems were proposed. Based on (R, Q) ordering policy, Mos-
tafa et al. [39] showed the information sharing’s influence on inventory costs. With the devel-
opment of meta-heuristic algorithms, several different meta heuristic algorithms have been
used to solve the two-tier supply chain problem [40], improve the efficiency of the supply
chain [41], solve the vehicle routing problem [42], and optimize the response time in medical
treatment [43].
The above methods confirm that ACO algorithm has the advantages of strong adaptability
and dynamic optimization through feedback mechanism. However, it also has the defects of
long convergence time and the inter-colony communication strategy is not delicate enough to
ensure the adaptability of ACO, so it is easy to fall into local optimization.
The basic idea of ACO algorithm is illustrated by an example of tourist route planning.
There are m scenic spots, dij(i, j = 1, 2. . ., m)is the distance between scenic spot i and j, τij(t)
represents the amount of information between scenic spot i and j, at time t, that is, the phero-
mone of simulated ants. Let there be n ants in total, pkij ðtÞ is the probability of ant k from scenic
spot i to j at time t, then Eq (1):
8 b
ta ðtÞZ ðtÞ
>
< P ij ij a b ; j 2 allowedk
t ðtÞZir ðtÞ
pkij ðtÞ ¼ g2allowedk ir
ð1Þ
>
:
0; otherwise

Among them, allowedk is the collection of scenic spots allowed to visit in the next step of
ant k. α represents the importance of pheromones on the path to the ants’ selection path, ηij is
the expected degree of transferred from scenic spot i to j, β is the degree of importance to ηij.
When β = 0, it is an heuristic algorithm representing positive feedback, and it is a greedy algo-
rithm in essence when α = 0. Generally, ηij is closely related to the distance between the two

PLOS ONE | https://doi.org/10.1371/journal.pone.0257317 September 16, 2021 4 / 16

PLOS ONE An improved ant colony optimization algorithm based on context for tourism route planning

scenic spots, generally ηij is calculated by Eq (2).

1
Zij ¼ ð2Þ
dij

Then, after each ant visits all the scenic spots, the pheromone is updated. The update rules
are as follows(Eqs (3), (4) and (5)):
tij ðt þ 1Þ ¼ rtij þ Dtij ðt; t þ 1Þ ð3Þ

X
n
Dtij ðt; t þ 1Þ ¼ Dtkij ðt; t þ 1Þ ð4Þ
k¼1

( Q
Lk
; when kth ant visit i from j
k
Dt ¼
ij ð5Þ
0; otherwise

Dtkij ðt; t þ 1Þ represents the amount of information that ant k stays on the road from scenic
spot i to j at the time of (t, t + 1), Δτij(t, t + 1) represents the increment of pheromone in scenic
spot i to j, and ρ is the change system of pheromone in the section with the value between
(0,1). Q is a constant, Lk is the length of the path taken by ant k.

Proposed method
Traditional tourism route planning methods introduce some heuristic algorithms to solve the
optimal path, and achieves good results. Huanwg introduced the chance algorithm into ACO,
which has good effect in dynamic tourism route planning [22]. Mei took time as the key con-
straint, and gave a tourism route planning prototype combined with ant colony algorithm
[44]. However, the above approaches suffer from two limitation when handling tourism route
planning. One limitation is that pheromone updating mechanism leads people to gather in a
scenic spot, which brings great challenges to tourism management and service. Another limita-
tion is that the accuracy of these approaches in large-scale tourism route planning problem is
not satisfactory. We propose a tourism route planning algorithm based on the integration of
environment feedback mechanism. The algorithm is based on ACO algorithm integrated con-
text aware(ACOCA) and introduces the contextual information feedback mechanism. It intro-
duces some contextual information such as comfort degree of the scenic spot, weather and
other contextual data into the algorithm, in order to improve the pheromone update method
to obtain the optimal tourism route on the basis of ensuring the user experience of tourists.
The main strategy of ACOCA algorithm: for the contextual information of the target scenic
spot, if it is beneficial to the tour, increase the pheromone concentration on the path to the sce-
nic spot, otherwise reduce it; At the same time, in order to avoid ant colony falling into local
optimization, the sub-path support degree of the sub-path is less than the threshold, the algo-
rithm will enter the next iteration.

Contextual information and representation

The contextual information of scenic spots, especially the weather and the number of tourists
significantly affect the willingness of tourists to travel. For weather information, different sce-
nic spots have different requirements for weather. For example, outdoor scenic spots are more
sensitive, while indoor tourism projects have lower requirements for weather. The factor

PLOS ONE | https://doi.org/10.1371/journal.pone.0257317 September 16, 2021 5 / 16

PLOS ONE An improved ant colony optimization algorithm based on context for tourism route planning

Weatheri of weather condition on scenic spot i, its vaule range is in the range of [0, 1], where 1
is strongly recommended and 0 is not recommended.
The number of tourists varies with the capacity of different scenic spots, so the comfort
degree of scenic spot is used to express the crowding intensity of scenic spot i is obtained by
Eq (6).
Comfi ¼ Ni =Nmax ð6Þ

Where Ni is the current tourists number of scenic spot i, Nmax is the maximum tourist number
scenic spot i can receive. Comfi is devided into four ranges, [0,0.25] means comfortable,
(0.25,0.50] stands for general comfortable, (0.50,0.75] represents average and (0.75,1] is
crowded. According to the management experience of the scenic spot, when the comfort
degree of the scenic spot is greater than 0.75, means that tourists are no longer recommended
to rush into the scenic spot.

Contextual distance
dij(i, j = 1, 2. . ., m) is the physical distance between the two scenic spots. When calculating the
physical distance between two scenic spots, the earth is calculated as an approximate sphere.
By obtaining the longitude and latitude coordinates of scenic spots, the calculation method of
distance is shown in Eq (7):
p p p
dij ¼ R � arccosðsinðLati � Þ � sinðLatj � Þ þ cosðLati � Þ
180 180 180
p p p ð7Þ
�cosðLatj � Þ � cosðjLoni Lonj j � ÞÞ �
180 180 180
where R is the radius of the earth, and the average radius is 6371.004km. The latitude and lon-
gitude coordinates of scenic spot i are expressed by (Loni, Lati).
It should be pointed out that considering the contextual information of the scenic spot, if
the number of tourists in scenic spot i is too large at a certain time, tourists are not recom-
mended to visit the scenic spot again; if the number of tourists in scenic spot j is appropriate,
tourists are recommended to visit the scenic spot j, then the contextual distance between scenic
spots Distanceij 6¼ Distanceji(Eq (8)).
8 pffiffiffiffiffiffiffiffi
< dij � Outi ; Comfj � 0:75
>
Distanceij ¼ qffiffiffiffiffiffiffiffi ð8Þ
>
: dij � Outi
; otherwise
Comfj

The outdegree Outi of departure scenic spot i is introduced to calculate the contextual dis-
tance. It means that when the destination scenic spot j is not suitable for sightseeing, the larger
of outdegree Outi, the higher probability of the scenic spot i to other scenic spots other than
the target scenic spot j.

Context feedback mechanism

Under the background of smart tourism construction, tourists can obtain the contextual data
of scenic spot in real time and make relevant preparations in advance. In ACO algorithm, the
only criterion for each ant to determine the path is pheromone concentration, but because of
the single criterion, it is easy to lead to uneven number of people between scenic spots. There-
fore, we introduce the combination of contextual data and pheromone as the criteria of ant
colony path selection. We propose an ant colony optimization integrated with context aware
algorithm, which is based on ACO algorithm and optimizes the pheromone updating rules

PLOS ONE | https://doi.org/10.1371/journal.pone.0257317 September 16, 2021 6 / 16

PLOS ONE An improved ant colony optimization algorithm based on context for tourism route planning

and transition probability of traditional ACO algorithm, so as to make the contextual informa-
tion of scenic spots participate in pheromone updating. The main principles of context infor-
mation collaborative improvement pheromone update are as follows:
1. When the weather and comfort degree of the scenic spot are not suitable for tourists, it is
necessary to artificially stimulate all route pheromones to the scenic spot according to the
contextual information of the scenic spot, and reduce the pheromone concentration of
these routes when updating the pheromones.
2. The pheromone update of basic ACO algorithm is positive feedback. Positive feedback is
conducive to the convergence of the algorithm, but it also reduces the diversity of the popu-
lation, resulting in the ant colony concentrated on a few routes, resulting in traffic pressure
and overload of scenic spots. In order to avoid the premature convergence of ant colony to
form the local optimal solution, the route selection probability is modified to make up for
the premature phenomenon caused by the disorder and irregular search of ACO algorithm.
Therefore, based on the above principle 1 and improved Eq (2), context distance is intro-
duced to replace physical distance(Eq (9)):
Zij ¼ Weatherj =Distanceij ð9Þ

In the Eq (9), the heuristic information of route(i, j) is determined by the weather, comfort
degree of scenic spot and physical distance dij. Positive feedback when the weather of the target
scenic spot j is not suitable for sightseeing; positive feedback when the comfort degree is suit-
able for sightseeing, otherwise negative feedback.
According to the above principle 2, because the pheromone updating strategy in basic ACO
algorithm only occurs on the optimal route, and the sub route of the optimal route may be lon-
ger, the ants choose the shortest route in the optimal route, which is the main reason for pre-
mature. The concept of subpath support degree(SPSD) is introduced, and the calculation
formula is shown in Eq (10). Among the optimal solutions found by each generation of ants,
the routes whose support degree of route(i, j) to the overall optimal path is greater than the
route contribution threshold are found, and the pheromones of route(i, j) are enhanced and
updated, and the update formula is shown in Eqs (11) and (12).
Distanceij
SPSDij ¼ ð10Þ
Li

( Q
Distanceij
; when SPSDij > x
0
Dtij ¼ ð11Þ
0; otherwise

tij ðt þ 1Þ ¼ Dtij0 þ tij ðt þ 1Þ ð12Þ

Where τij(t + 1) is still calculated by Eq (3), x is the route support threshold, when SPSDij
is greater than the threshold, the pheromone of the sub route(i, j) is updated, and Q is the
pheromone increase coefficient, that is, the total amount of pheromones released by ants in an
iteration.

ACOCA algorithm description

The basic goal of ACOCA algorithm is that the planned route is an optimal solution: firstly,
the proportion of comfort passing through the scenic spots is low, and secondly, the path

PLOS ONE | https://doi.org/10.1371/journal.pone.0257317 September 16, 2021 7 / 16

PLOS ONE An improved ant colony optimization algorithm based on context for tourism route planning

distance is the shortest. According to the contextual information of scenic spots and path dis-
tance, the cost function is obtained, as shown in Eq (13), where makes the path cost V the low-
est, sij is the comfort degree of the scenic spot i to j, and dij is the distance from scenic spot i to
j. The constraints are expressed by Eqs (14)–(17), Where skij and yki are calculated by Eqs (18)
and (19).
X
m X
n X
n
minV ¼ sij dij ð13Þ
k¼1 j¼1 i¼1

X
m
s:t yki ¼ 1; i ¼ 2; 3; :::; n ð14Þ
k¼1

X
n
skij ¼ yki ; j ¼ 2; 3; :::; n; k ¼ 1; 2; :::; m ð15Þ
i¼2

X
n
skij ¼ yki ; i ¼ 2; 3; :::; n; k ¼ 1; 2; :::; m ð16Þ
j¼2

X
n
yki ci � Cmax ; k ¼ 1; 2; :::; m ð17Þ
i¼2

(
1; user k from scenic spot i to j
skij ¼ ; i; j ¼ 2; 3; :::; n; k ¼ 1; 2; :::; m ð18Þ
0; otherwise

(
1; scenic spot i meets user k
yki ¼ ; i ¼ 2; 3; :::; n; k ¼ 1; 2; :::; m ð19Þ
0; otherwise

The pheromone is updated through the contextual information of scenic spots, the basic
workflow as following(Fig 1).

Experimental evaluation
Datasets and parameter settings
In order to test the performance of ACOCA algorithm in route planning in different regions,
we adopt three groups of scenic spots with different geographical distribution, as shown in Fig
2 below.
All of the algorithms implement in Python, we run the proposed algorithm and the base
algorithm 10 times in each dataset, and calculate the average of the experimental results for
comparison. The parameters settings in our experiments of ACOCA as Table 1, and the exper-
imental environment shown as Table 2.

Evaluation protocol
We mainly evaluate from the following three aspects: total route distance, convergence time
and user comfort ratio.
Total route distance: From the perspective of economic cost, the total distance of the route
is evaluated. The tourism practice route planning algorithm plans the total distance of the

PLOS ONE | https://doi.org/10.1371/journal.pone.0257317 September 16, 2021 8 / 16

PLOS ONE An improved ant colony optimization algorithm based on context for tourism route planning

Fig 1. The workflow of ACOCA algorithm. Input: set the parameters of the algorithm, such as the number of ants,
the maximum number of iterations, the initial comfort degree of each scenic spot, and the local optimal threshold x.
Output: the optimal solution and its route set. Step 1: randomly put the ants in different scenic spots as the starting
point; Step 2: each ant calculates the solution space and the comfort degree of each scenic spot; Step 3: calculates and
updates the pheromone concentration on the connection path between scenic spots according to Eqs (5) and (9); Step
4: updates the comfort degree of the scenic spot every iteration; Step 5: judge whether the algorithm falls into local
optimum. Judgment method: compare the global optimal solution obtained in this iteration with the global optimal
solution obtained in the previous iteration. If the continuous x-times optimal solution is not improved, it is considered
to fall into local optimum and implements the local optimal strategy; otherwise, it enters step 6; Step 6: determines
whether the maximum number of iterations is reached. If so, the algorithm ends, otherwise, step 1 is executed again.
https://doi.org/10.1371/journal.pone.0257317.g001

route from the starting point to the starting point after visiting several scenic spots in a certain
order.
Convergence time: From the point of view of the performance of the algorithm, the conver-
gence time is evaluated. Because ACOCA and the base algorithms are heuristic algorithms, the
algorithm takes a finite number of iterations to get a stable solution, that is, the convergence
time represents the performance of the algorithm.
User comfort ratio: From the perspective of tourist user experience, the scenic comfort
degree is evaluated. It mainly measures the tourists’ comfort in the planned route, such as the
scenic spot is crowded or not, the weather of the scenic spot is suitable for visiting or not. The
evaluation is based on the number of crowded scenic spots in the whole journey.

Experimental results
The running results of ACOCA algorithm and base algorithms in three groups of scenic spots
with different geographical distribution are given in Table 3. The results are the average of
each algorithm running 10 times in each group of data. The comfort ratio is the percentage of
scenic spots with comfort degree higher than 0.75 in all scenic spots of the planned route(the
higher the comfort ratio, the lower the overall user satisfaction).

PLOS ONE | https://doi.org/10.1371/journal.pone.0257317 September 16, 2021 9 / 16

PLOS ONE An improved ant colony optimization algorithm based on context for tourism route planning

Fig 2. Scenic spots. A: scenic spots across different cities, including 15 scenic spots in Shanghai, Hangzhou and
Nanjing of China. B: scenic spots in the same city, including 15 scenic spots in Shanghai. C: different scenic spots in
one scenic spot, we take 15 scenic spots of Qiandao Lake.
https://doi.org/10.1371/journal.pone.0257317.g002

PLOS ONE | https://doi.org/10.1371/journal.pone.0257317 September 16, 2021 10 / 16

PLOS ONE An improved ant colony optimization algorithm based on context for tourism route planning

Table 1. Parameters settings.

Ant Numbers Maximum Iterations Initial Pheromone α β ρ
20 100 0.5 1.0 4.5 0.1
https://doi.org/10.1371/journal.pone.0257317.t001

Table 2. Experimental environment.

hardware and software values
CPU Intel i5 @ 2.4GHz
RAM 8GB
OS Windows 10
IDE PyCharm in Python 3.7
https://doi.org/10.1371/journal.pone.0257317.t002

It can be seen from Figs 3 and 4 that the comfort ratio is obviously better than the ACO
algorithm by introducing the comfort degree and weather information, which improves the
user’s tourism experience. Through real-time contextual information, it can help users to plan
tourism routes better. When the distance between scenic spots becomes larger, the optimal
path of ACOCA algorithm has some advantages over ACO algorithm. By introducing the sup-
port degree of path, the local optimal problem is effectively solved, and the solution of the
problem has a good performance. Therefore, on the premise that the efficiency of the algo-
rithm is almost the same, ACOCA algorithm can better improve the tourists’ travel experience.
At the same time, the distance of the travel route planned by ACOCA algorithm is also better
than that of ACO algorithm.

Table 3. The results of experiments.

Different Dataset Average Distance Average Convergence Time Average Comfort Ratio
ACOCA(same city) 203.4739 2.2676 26%
ACOCA(different cities) 885.7403 2.3338 27%
ACOCA(same scenic spot) 70.6389 2.3286 28%
ACO(same city) 205.7595 3.0594 73%
ACO(different cities) 892.3387 2.5736 88%
ACO(same scenic spot) 71.7929 3.0335 78%
https://doi.org/10.1371/journal.pone.0257317.t003

Fig 3. Performance of ACOCA and ACO. From the comparison results of the two algorithms in different datasets,
the distance of ACOCA algorithm is better than that of ACO algorithm.
https://doi.org/10.1371/journal.pone.0257317.g003

PLOS ONE | https://doi.org/10.1371/journal.pone.0257317 September 16, 2021 11 / 16

PLOS ONE An improved ant colony optimization algorithm based on context for tourism route planning

Fig 4. Comparison of two algorithms in convergence time. From the comparison results of the two algorithms in
different datasets, the convergence time of ACOCA algorithm is better than that of ACO algorithm.
https://doi.org/10.1371/journal.pone.0257317.g004

Fig 5. ACOCA algorithm in the same city planning results and convergence. A: The convergence of ACOCA
algorithm. B: The result of ACOCA algorithm.
https://doi.org/10.1371/journal.pone.0257317.g005

Taking the route planning results of ACOCA algorithm in the same city as an example, Fig
5 shows the result of ACOCA algorithm and its convergence. The results show that the algo-
rithm is stable after the 40th iteration, and the convergence effect is good, and the planning
route is in line with the actual context.
At the same time, in order to verify the universality of ACOCA algorithm, experiments
are carried out on the open dataset TSPLIB st70, and the experimental results are shown in
Table 4.

Table 4. The results of experiments.

Algorithm Distance(st70) Time(st70) Distance(different cities) Time(different cities)
ACO 767.7446 94.1582 892.3387 2.5736
GA 979.5677 20.1540 897.3524 1.2420
Hybrid-ACO 884.2793 95.6306 976.4651 4.2887
ACOCA 703.0122 75.3935 885.7403 2.3338
https://doi.org/10.1371/journal.pone.0257317.t004

PLOS ONE | https://doi.org/10.1371/journal.pone.0257317 September 16, 2021 12 / 16

PLOS ONE An improved ant colony optimization algorithm based on context for tourism route planning

It can be seen from Table 4 that the ACOCA algorithm proposed in this paper has great
advantages in solving the distance of the optimal path on two datasets, especially in the large-
scale dataset st70, the ACOCA algorithm is 9.21% shorter than the ant colony algorithm,
25.78% shorter than the Hybrid-ACO algorithm, and has obvious advantages in planning the
shortest path distance. But genetic algorithm (GA) is a global solution, its convergence time is
incomparable to ACO algorithm. In addition, the four algorithms in the same city of 15 scenic
spots path planning, ACOCA algorithm in the planning path distance is also the best, and the
convergence time is not different from GA algorithm, indicating that ACOCA algorithm per-
forms better in small-scale datasets.

Conclusion and future work

In order to optimize the tourism route planning problem, in this paper, an improved ant col-
ony optimization(ACOCA) algorithm based on context-aware mechanism, pheromone updat-
ing strategy is proposed to balance route distance and user comfort degree. In the proposed
ACOCA algorithm, we introduce the weather and comfort degree of scenic spots to decide
the pheromone updating strategy, and avoid to fall into the local optimum value, the concept
of sub-path support degree was set. In order to verify the optimization performance of the
ACOCA algorithm, We select three groups of scenic spots with different regional ranges, each
group includes 15 scenic spots, and make route planning for each group of scenic spots; At the
same time, we also carried out simulation experiments on TSPLIB dataset. The experimental
results show that the performance of ACOCA algorithm is compared with basic ACO algo-
rithm, genetic algorithm and Hybrid-ACO algorithm on the two datasets, and its operation
results are optimized in route distance, convergence time and user comfort ratio. In particular,
compared with Hybrid-ACO algorithm, the travel route distance is reduced by 20.5% and the
convergence time is reduced by 21.2%, At the same time, user travel comfort has also been
greatly improved. ACOCA algorithm has better optimization ability and user comfort than the
baseline algorithms.
However, the efficiency of ACOCA algorithm needs to be further improved when the scale
of scenic spots increases greatly. At the same time, we plan to include another contextual fac-
tors such as tourism cost into the algorthm. In the future work, the ACOCA algorithm will be
studied deeply.

Acknowledgments
The author would like to thank the editor and the anonymous reviewers for their valuable
comments.

Author Contributions
Conceptualization: Shengbin Liang.
Data curation: Tongtong Jiao.
Formal analysis: Wencai Du, Shenming Qu.
Funding acquisition: Shengbin Liang, Wencai Du.
Investigation: Wencai Du.
Methodology: Shengbin Liang.
Project administration: Tongtong Jiao.
Resources: Tongtong Jiao.

PLOS ONE | https://doi.org/10.1371/journal.pone.0257317 September 16, 2021 13 / 16

PLOS ONE An improved ant colony optimization algorithm based on context for tourism route planning

Software: Shengbin Liang, Tongtong Jiao, Shenming Qu.

Supervision: Shengbin Liang, Tongtong Jiao, Shenming Qu.
Validation: Tongtong Jiao.
Visualization: Tongtong Jiao.
Writing – original draft: Shengbin Liang.
Writing – review & editing: Tongtong Jiao.

References
1. Dorigo M, Maniezzo V, Colorni A. Distributed optimization by ant colonies. In: European Conference on
Artificial Life. vol. 134; 1991. p. 134–142.
2. Stützle T, Hoos H. Improvements on the ant-system: Introducing the max-min ant system. In: Artificial
neural nets and genetic algorithms. Springer; 1998. p. 245–249.
3. Yang K, You X, Liu S, Pan H. A novel ant colony optimization based on game for traveling salesman
problem. Applied Intelligence. 2020; 50(12):4529–4542. https://doi.org/10.1007/s10489-020-01799-w
4. Deng W, Xu J, Zhao H. An improved ant colony optimization algorithm based on hybrid strategies for
scheduling problem. IEEE access. 2019; 7:20281–20292. https://doi.org/10.1109/ACCESS.2019.
2897580
5. Qian X, Zhong X. Optimal individualized multimedia tourism route planning based on ant colony algo-
rithms and large data hidden mining. Multimedia Tools and Applications. 2019; 78(15):22099–22108.
https://doi.org/10.1007/s11042-019-7537-0
6. Liang Y, Wang L. Applying genetic algorithm and ant colony optimization algorithm into marine investi-
gation path planning model. Soft Computing. 2020; 24(11):8199–8210. https://doi.org/10.1007/s00500-
019-04414-4
7. Che G, Liu L, Yu Z. An improved ant colony optimization algorithm based on particle swarm optimization
algorithm for path planning of autonomous underwater vehicle. Journal of Ambient Intelligence and
Humanized Computing. 2020; 11(8):3349–3354. https://doi.org/10.1007/s12652-019-01531-8
8. Wang P, Bai J, Meng J. A hybrid genetic ant colony optimization algorithm with an embedded cloud
model for continuous optimization. Journal of Information Processing Systems. 2020; 16(5):1169–
1182.
9. Han N, Wang N, Tian N. Ant colony optimization for assembly sequence planning based on parameters
optimization. Frontiers of Mechanical Engineering. 2021; 16(2):393–409. https://doi.org/10.1007/
s11465-020-0613-3
10. Ma W, Zhou X, Zhu H, Li L, Jiao L. A Two-stage Hybrid Ant Colony Optimization for High-Dimensional
Feature Selection. Pattern Recognition. 2021; 116(1):107933. https://doi.org/10.1016/j.patcog.2021.
107933
11. Akbarpour N, Salehi-Amiri A, Hajiaghaei-Keshteli M, Oliva D. An innovative waste management system
in a smart city under stochastic optimization using vehicle routing problem. Soft Computing. 2021; 25
(8):6707–6727. https://doi.org/10.1007/s00500-021-05669-6
12. Liu Y, Cao B, Li H. Improving ant colony optimization algorithm with epsilon greedy and Levy flight.
Complex and Intelligent Systems; p. 1–12.
13. Qi X, Gan Z, Liu C, Xu Z, Ouyang C. Collective intelligence evolution using ant colony optimization and
neural networks. Neural Computing and Applications. 2021;(1).
14. Lu Y, Benlic U, Wu Q. A highly effective hybrid evolutionary algorithm for the covering salesman prob-
lem. Information Sciences. 2021; 564:144–162. https://doi.org/10.1016/j.ins.2021.02.053
15. Haroun SA, Jamal B, et al. A performance comparison of GA and ACO applied to TSP. International
Journal of Computer Applications. 2015; 117(20):28–35. https://doi.org/10.5120/20674-3466
16. Ahmed ZH. Genetic algorithm for the traveling salesman problem using sequential constructive cross-
over operator. International Journal of Biometrics & Bioinformatics (IJBB). 2010; 3(6):96–105.
17. Liu Z, Huang L. A mixed discrete particle swarm optimization for TSP. In: 2010 3rd International Confer-
ence on Advanced Computer Theory and Engineering (ICACTE). vol. 2. IEEE; 2010. p. V2–208.
18. Hao M, Xin G. UAV route planning based on the genetic simulated annealing algorithm. In: International
Conference on Mechatronics and Automation; 2010.

PLOS ONE | https://doi.org/10.1371/journal.pone.0257317 September 16, 2021 14 / 16

PLOS ONE An improved ant colony optimization algorithm based on context for tourism route planning

19. Akka K, Khaber F. Mobile robot path planning using an improved ant colony optimization. International
Journal of Advanced Robotic Systems. 2018; 15(3):1–7. https://doi.org/10.1177/1729881418774673
20. Maheshwari P, Sharma AK, Verma K. Energy efficient cluster based routing protocol for WSN using but-
terfly optimization algorithm and ant colony optimization. Ad Hoc Networks. 2021; 110:102317. https://
doi.org/10.1016/j.adhoc.2020.102317
21. Feng YJ, Zhang X, Liu XH. Research on The Optimization of The Enterprise Logistics Distribution
Based on ACO Algorithm. In: Advanced Materials Research. vol. 912. Trans Tech Publ; 2014. p.
1900–1903.
22. Huang HC. The application of ant colony optimization algorithm in tour route planning. Journal of Theo-
retical and Applied Information Technology. 2013; 52(3):343–347.
23. Yong W. Hybrid max–min ant system with four vertices and three lines inequality for traveling salesman
problem. Soft Computing. 2015; 19(3):585–596. https://doi.org/10.1007/s00500-014-1279-8
24. Yang Z, Fang L, Zhang X, Zuo HJO, Engineering Li. Controlling a scattered field output of light passing
through turbid medium using an improved ant colony optimization algorithm. 2021; 144:106646.
25. Jia ZH, Cui YF, Li KJAI. An ant colony-based algorithm for integrated scheduling on batch machines
with non-identical capacities. 2021; p. 1–18.
26. Wu H, Gao Y, Wang W, Zhang Z. A hybrid ant colony algorithm based on multiple strategies for the
vehicle routing problem with time windows. Complex & Intelligent Systems. 2021; p. 1–18.
27. Miao C, Chen G, Yan C, Computers YJ, Engineering I. Path planning optimization of indoor mobile
robot based on adaptive ant colony algorithm. 2021; p. 107230.
28. Yu J, You X, Liu SJSC. Ant colony algorithm based on magnetic neighborhood and filtering recommen-
dation. 2021; p. 1–16.
29. Cerda J, Rojas-Morales N, Minutolo MC, Kristjanpoller W. High Frequency and Dynamic Pairs Trading
with Ant Colony Optimization. 2021;.
30. Zhang S, Wang X, Gao J, Wu X, Liu JJJoEE, Technology. SVPWM Overmodulation Region II Control
Method Based on the Chaos Ant Colony Algorithm. 2021;(6).
31. Yu J, You X, Liu S. Dynamic reproductive ant colony algorithm based on piecewise clustering. Applied
Intelligence. 2021;(9).
32. Xiao Z, Wang Z, Liu D, Wang HJJoIM. A path planning algorithm for PCB surface quality automatic
inspection. 2021; p. 1–13.
33. Wang Y. Application of ant colony algorithm and artificial intelligence in training simulation of athletes in
sports arena. Journal of Ambient Intelligence and Humanized Computing. 2021;. https://doi.org/10.
1007/s12652-021-03212-x
34. Kanso B, Kansou A, Yassine AJRRO. Open Capacitated ARC routing problem by Hybridized Ant Col-
ony Algorithm. 2021; 55:639.
35. Shi B, Zhang YJE. A novel algorithm to optimize the energy consumption using IoT and based on Ant
Colony Algorithm. 2021; 14(6):1709.
36. Axsäter S. Simple Solution Procedures for a Class of Two-Echelon Inventory Problems. 1990; 38(No
1):64–69.
37. Axster S. Exact and Approximate Evaluation of Batch-Ordering Policies for Two-Level Inventory Sys-
tems. Operations Research. 1993; 41.
38. Hajiaghaei-Keshteli M, Sajadifar SM. Deriving the cost function for a class of three-echelon inventory
system with N-retailers and one-for-one ordering policy. International Journal of Advanced Manufactur-
ing Technology. 2010; 50(1):343–351. https://doi.org/10.1007/s00170-009-2486-9
39. Hajiaghaei-Keshteli M, Sajadifar SM, Haji R. Determination of the economical policy of a three-eche-
lon inventory system with (R, Q) ordering policy and information sharing. International Journal of
Advanced Manufacturing Technology. 2011; 55(5-8):831–841. https://doi.org/10.1007/s00170-010-
3112-6
40. Amiri S, Zahedi A, Kazemi M, Soroor J, Hajiaghaei-Keshteli M. Determination of the Optimal Sales
Level of Perishable Goods in a Two-echelon Supply Chain Network. Computers and Industrial Engi-
neering. 2019; 139:106156. https://doi.org/10.1016/j.cie.2019.106156
41. Abdi A, Abdi A, Akbarpour N, Amiri AS, Hajiaghaei-Keshteli M. Innovative approaches to design and
address green supply chain network with simultaneous pick-up and split delivery. Journal of Cleaner
Production. 2020; 250(Mar.20):119437.1–119437.19.
42. Guan B, Zhao Y, Li Y. An improved ant colony optimization with an automatic updating mechanism for
constraint satisfaction problems. Expert Systems with Applications. 2021; 164:114021. https://doi.org/
10.1016/j.eswa.2020.114021

PLOS ONE | https://doi.org/10.1371/journal.pone.0257317 September 16, 2021 15 / 16

PLOS ONE An improved ant colony optimization algorithm based on context for tourism route planning

43. Zahedi A, Salehi-Amiri A, Smith NR, Hajiaghaei-Keshteli M. Utilizing IoT to design a relief supply chain
network for the SARS-COV-2 pandemic. Applied Soft Computing. 2021; 104(5–8):107210. https://doi.
org/10.1016/j.asoc.2021.107210 PMID: 33642961
44. Mei Y. Study on the application and improvement of ant colony algorithm in terminal tour route planning
under Android platform. Journal of Intelligent & Fuzzy Systems. 2018; 35(3):2761–2768. https://doi.org/
10.3233/JIFS-169628

PLOS ONE | https://doi.org/10.1371/journal.pone.0257317 September 16, 2021 16 / 16