The Investigation of Advancements in Intelligent Tourism Route

Planning Based on Path Generation Algorithms

Shiyi Li

and Yan Yan

Intelligent Science and Technology, Fudan University, Shanghai, China

Logistics Management (Operation Management and Business Analysis),

Southwestern University of Finance and Economics, Chengdu, China

Keywords: Path Planning, Heuristic Algorithm, Collaborative Filtering Algorithm, Genetic Algorithm.

Abstract: Due to individual differences in preferences and travel costs, it is challenging to provide accurate travel route

planning to allow users to obtain personalized travel route recommendations. In this article, we analyze in

detail the four commonly used path generation methods in the past, namely exact algorithm, heuristic

algorithm, collaborative filtering algorithm and genetic algorithm, and analyze the ideas and specific

implementation of these algorithms. We then discuss the practical application of tourism route planning, The

A* algorithm and genetic algorithm, which are heuristic algorithms, are widely used in navigation application

design, travel planning, and other path planning problems. The collaborative filtering algorithm is used to

implement personalized route recommendations based on user preferences. Finally, we come to the

conclusion, this study can sort out and integrate the research results in the field of tourism route planning,

helping researchers understand the development status and trends of current research. With the development

of technologies such as artificial intelligence and data mining, future travel route planning may become

increasingly intelligent, capable of making personalized recommendations based on tourists' preferences, real-

time traffic and other factors.

1 INTRODUCTION

Path planning refers to the process of determining,

within a particular environment, the optimal route

from a starting point to an end point (Damos, 2021).

As the economic developing and living standard

rising, traveling has become a significant kind of

recreation. It's now crucial to design their own route

for individuals in advance of travel in order to ensure

a nice experience. The qualities of a self-driving tour

are independence, adaptability, diversity, and

selection. The self-driving tour's path planning, which

can be time-consuming and expensive, greatly affects

the entire experience. To enhance the self-driving trip

experience and lower overall costs, we should

investigate a self-driving travel path planning

technique.

The predecessor of path planning problems is the

Traveling Salesman Problem (TSP). Traditional

methods for solving this kind of problem involve

https://orcid.org/0009-0001-5640-3373

https://orcid.org/0009-0007-2036-509X

graph theory algorithms, such as using the Dijkstra

algorithm to process weighted directed graphs to

obtain the shortest path for the urban transportation

(Sari, 2021). Another approach is heuristic algorithms

， such as the ant colony algorithm and genetic

algorithms. Adding more influencing factors to those

traditional algorithms is the main way to improve

model performance. For example, the Adaptive Ant

Colony Algorithm (IAACO) introduces angle

guidance factors into the ACO to effectively improve

the real-time performance of the model (Miao, 2021).

In the case of self-driving tour path planning, it is

essential to handle larger scales and varieties of data

and provide users with comfortable and personalized

customized routes. However, traditional algorithms

have difficulties in solving this issue.

Consequently, a lot of research has been done in

the area of trip route planning in recent years using

machine learning and deep learning models. Based on

autoregressive time series models and deep learning

Li, S. and Yan, Y.

The Investigation of Advancements in Intelligent Tourism Route Planning Based on Path Generation Algorithms.

DOI: 10.5220/0012956700004508

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Engineering Management, Information Technology and Intelligence (EMITI 2024), pages 485-489

ISBN: 978-989-758-713-9

485

models, the new model can dynamically adjust based

on real-time traffic locations and interact with the

environment to achieve real-time traffic data

prediction, minimizing travel time (Geng, 2021).

Using techniques for heterogeneous network

embedding to handle multimodal input and extract

valuable information. After then, by integrating

supplementary data using attention-based deep

learning to forecast the subsequent Points of Interest

(POI), achieving a multi-task path planning model

(Huang, 2023). Green Path incorporates

environmental factors into the model consideration,

introduces the environmental impedance model,

offers walking and cycling as travel mode choices,

and can optimize route planning results based on

different environmental factors, achieving a

multimodal route planning model (Helle, 2023). This

article aims to summarize the research work on travel

path planning problems, identify unresolved issues in

this field, and provide personal insights for future

work directions.

The paper is structured as follows. First, we

introduce some the previous traditional path planning

algorithms. Then, we discuss the improvements on

the previous algorithm in detail, the feasibility and the

specific implementation of tourism route planning,

and finally summarize the entire article and draw the

final conclusion.

2 METHODS

2.1 Exact Algorithm

Exact algorithms refer to algorithms that obtain the

global optimal solution by calculating all possible

solutions, including Branch and Bound, Cutting Plane,

Branch and Cut Methods and etc. The Branch and

Bound Method divides the problem into multiple

subproblems, determines the upper and lower bounds

of each subproblems, solves the local optimal

solution, and then further branches to bound. Based

on this method, Christofides et al. utilized the

spanning, tree search method to accurately solve

problems with up to 25 customer points (Christofides,

1981). The maximum clique problem proposed by

Jiang et al. executes multiple subtasks in parallel and

allows the subtasks to share the latest computed lower

bounds, when a subtask's upper bound is smaller than

the maximal clique's lower bound, the subtask is

terminated to narrow the search scope, effectively

improving the efficiency of the algorithm (Jiang,

2023).

2.2 Heuristic Algorithm

Heuristic algorithm is an effective method derived

from biological activities in nature, for solving path

planning problems. Line and Kernighan utilized this

approach to deal with the classical Travelling

Salesman Problem (Lin, 1973). Heuristic algorithm

typically cannot guarantee to obtain the global

optimal solution, they can find relatively good

approximate solutions within a specified time frame,

making them capable of handling large-scale NP hard

problems. One of the most used heuristic algorithms

is the Ant Colony Optimization Algorithm. The Ant

Colony Optimization Algorithm is modeled after the

foraging behaviors of real ants. Ants mark the most

efficient paths for other members to follow by

releasing pheromones (Dorigo, 2006). In order to

increase the size of the solution space and decrease

the impact of a small number of meeting ants on the

results, Gao suggested merging searching ants and

adding a threshold constant, effectively optimizing

the slow convergence and low efficiency of the

original ant colony algorithm (Gao, 2020). Li et al.

merged the ant colony method with the greedy

algorithm to create a novel ant colony optimization

technique based on an adaptable greedy approach.

Greedy algorithm chooses the local best solution at

each stage. By modifying control parameters, the

greedy algorithm is used to continuously change the

ant colony's preference degree for route selection,

accelerating the iteration process and hastening the

discovery of the global optimal solution (Li, 2022).

2.3 Collaborative Filtering Algorithm

The idea behind collaborative filtering is to identify

some similarities (either between users or similarity

between objects) through the behavior of the group,

and use these similarities to make decisions and

recommendations for users. In order to achieve

customized travel route planning, many studies have

developed a system that combines big data, satellite

positioning, crawler algorithms and other

technologies. Cenamor studied and discussed travel

routes based on positioning situations and developed

Plantour (Cenamor, 2013), a system for generating

routes and recommending tourist attractions. Based

on user status, user-contributed material and other

information from social media accounts, the Plantour

system will collect and compile large amounts of data

and determine which attractions the user likes and

dislikes. Depending on the user's trip time and the

most well-liked attractions on the Internet, to offer

menu-based preset routes and attraction suggestions.

EMITI 2024 - International Conference on Engineering Management, Information Technology and Intelligence

486

2.4 Genetic Algorithm

Simple Genetic Algorithm (SGA) generates an initial

population through a random method. The poor

individual fitness of the beginning population will

limit the algorithm's pace of convergence to some

degree. Yu studied the genetic algorithm in light of

this (Yu, 2014). The heuristic crossover operator,

which is based on the greedy approach, is utilized to

maximize the crossover results. The greedy algorithm

is used to start the population on the basis of the basic

genetic algorithm. Studies and research demonstrate

that when the population size is modest, the enhanced

genetic algorithm may be applied to achieve more

dependable optimization capabilities.

An enhanced single-parent genetic algorithm was

suggested by Hu et al (Hu, 2019). This paper

developed a novel coding technique that, during

population initialization, may produce people with

random distribution centers by drawing inspiration

from the two-stage chromosomal coding method:

utilizing an enhanced Lastly, a mixed selection

operator is employed to retain and select the

population, keeping the algorithm from prematurely

convergent. The single-parent genetic operation is

utilized to optimize the route. According to the study,

the algorithm may be used to design trip routes and

find the shortest path with high performance.

In order to address the traveling salesman issue,

Chen et al. enhanced the evolutionary method using

pointer networks (Chen, 2020), a neural network that,

given a discrete input sequence, can ascertain the

conditional probability of an output sequence. Low-

and medium-level problems can be solved by it

successfully. High accuracy predictions may be made

on the solution to dimensional combination

optimization issues. The pointer network's basic idea

is to translate an input into a sequence of pointers that,

in accordance with probability, refer to the input

sequence elements. The improved genetic algorithm

will generate a high-quality initial population through

the pointer network. The initial population of the

pointer network is then combined with the random

initial population to determine the optimal individual.

The retention strategy retains excellent individuals on

both sides to form a new initial population. According

to experimental findings, the enhanced algorithm's

convergence speed and optimization capacity have

greatly increased. As a consequence, it is possible to

use genetic algorithms more successfully to the

traveling salesman issue.

3 DISCUSSIONS

3.1 The Application of Path Planning

Algorithms

The heuristic algorithms, collaborative filtering

algorithms, and genetic algorithms mentioned above

have demonstrated excellent performance in solving

various practical path planning problems. Travel apps

such as Ctrip, Gaode Maps, Uber, etc., have utilized

these algorithms to provide users convenient and

efficient services for everyday needs and travel

planning.

The A* algorithm, a heuristic algorithm widely

used in path planning, exhibits significant advantages

in practical scenarios due to its efficiency,

completeness, and robust adaptability. Navigation

systems, autonomous vehicle driving, and robot path

planning widely employ the A* algorithm. From our

perspective, the popularity of the A* algorithm stems

primarily from its ability to efficiently find the

optimal path while maintaining strong real-time

performance and adaptability. During the search

process, the A* algorithm utilizes heuristic

information to effectively reduce the search space,

thereby improving search efficiency. For instance,

Gaode Maps employs the A* algorithm combined

with straight-line distance estimation and cost

function design, comprehensively considering factors

such as road length, traffic conditions, and congestion,

to give consumers the best possible mix of path

planning from the starting point to the destination as

fast as feasible, and can provide users with better

solutions based on real-time conditions.

The genetic algorithm, which has strong global

search capabilities, is suitable for solving multimodal

optimization problems. In our view, the advantages of

genetic algorithms lie in their parallelism, ability to

handle high-dimensional, nonlinear, and non-

differentiable optimization problems, and strong

adaptability. They can also perform parallel

computations in distributed environments, exhibiting

good scalability and generalization ability to better

handle practical path planning problems with massive

data. Shivgan effectively solved the drone path

planning problem using genetic algorithms (Shivgan,

2020). Compared to greedy algorithms, genetic

algorithms are less likely to fall into local optima,

significantly reducing energy consumption and

improving the performance and efficiency of drone

operations.

Collaborative filtering algorithms enable path

planning models to better address user preference

issues and achieve personalized recommendations.

The Investigation of Advancements in Intelligent Tourism Route Planning Based on Path Generation Algorithms

487

These algorithms utilize large amounts of user data

and form information to discover user preferences

and historical behavior, thereby customizing the

optimal travel itinerary tailored to users' habits and

preferences. For example, the Plantour system is

designed based on collaborative filtering algorithms.

While generating routes, it provides personalized

recommendation services to users based on the

analysis of a large amount of data (Cenamor, 2013).

3.2 Disadvantages and Challenges

Although the aforementioned three algorithms have

achieved remarkable success in practical applications,

each method still exhibits inevitable limitations.

Firstly, the A* algorithm heavily relies on heuristic

information to guide the search process. The heuristic

function chosen has a major influence on how

accurate the model's output is. However, in complex

and dynamic environments, heuristic information is

often incomplete or inaccurate, which may hinder the

A* algorithm from finding the globally optimal

solution. Secondly, data, as a valuable resource in

today's society, is not always available. This limits the

effectiveness of models such as collaborative filtering

algorithms, which rely on capturing data features

from large amounts of information, leading to

inaccurate recommendations. Lastly, genetic

algorithms, with good scalability and adaptability,

typically require extensive iterations and adjustments

to find the optimal solution for multimodal

optimization problems in the search space,

significantly increasing computational costs.

3.3 Future Prospects

In light of the aforementioned issues, several

potential solutions could be determined. Firstly, one

could adopt a hybrid approach that combines multiple

algorithms to mitigate the shortcomings of individual

methods and amplify their strengths. For example,

integrating the A* algorithm with reinforcement

learning techniques could be advantageous. By

leveraging reinforcement learning's ability to

autonomously learn from interactions with the

environment, it reduces reliance on heuristic

functions, adapts better to dynamic changes, and

enhances the robustness and stability of path planning.

Secondly, incorporating meta-learning can reduce the

model's dependence on data, enabling it to quickly

infer the optimal learning strategy for new tasks with

limited data. Lastly, by utilizing parallelization

techniques or computational platforms such as

Apache Spark, one can accelerate the search process,

reduce iteration counts, simplify genetic algorithms,

enhance algorithm efficiency, and arrive at a more

affordable ideal option.

4 CONCLUSIONS

In this work, we offer a summary associated with

tourist route planning and its four commonly used

algorithms. This paper also discusses the practical

application of tourism route planning with a large

number of examples. Analyzing the limitations of the

above four algorithms in solving real-world problems

today, and proposed personal insights on how to

further optimize and improve existing models by

integrating cutting-edge technologies such as deep

learning and machine learning. This study

systematically summarizes the current research

progress in terms of designing tourist routes,

including development of different methods,

algorithms and application scenarios, which can help

researchers understand the latest developments in this

field. This study also conducts an in-depth analysis of

the existing problems and limitations of current

research, proposes the focus and direction of future

research, and provides guidance and inspiration for

researchers. Future travel route planning will

increasingly incorporate big data and artificial

intelligence technology to achieve in-depth

exploration of tourist behaviors and preferences,

providing better support for personalized travel route

planning.

AUTHORS CONTRIBUTION

All the authors contributed equally, and their names

were listed in alphabetical order.

REFERENCES

Cenamor, I., De la Rosa, T., & Borrajo, D. 2013. OnDroad

Planner: Building Tourist Plans Using Traveling Social

Network Information. AAAI Workshop - Technical

Report, 1, 14-15.

Chen, S., Lin, P., & Huang, P. 2020. Pointer Network

Improved Genetic Algorithm for Solving Traveling

Salesman Problem. Computer Engineering and

Applications, 56(19), 231-236.

Christofides, N., Mingozzi, A., & Toth, P. 1981. Exact

algorithms for the vehicle routing problem, based on

EMITI 2024 - International Conference on Engineering Management, Information Technology and Intelligence

488

spanning tree and shortest path relaxations.

Mathematical Programming, 20, 255-282.

Damos, M.A., Zhu, J., Li, W., Hassan, A., & Khalifa, E.

2021. A Novel Urban Tourism Path Planning Approach

Based on a Multiobjective Genetic Algorithm. ISPRS

International Journal of Geo-Information, 10(8), 530.

Dorigo, M., Birattari, M., & Stutzle, T. 2006. Ant colony

optimization. IEEE Computational Intelligence

Magazine, 1(4), 28-39.

Gao, W. 2020. New ant colony optimization algorithm for

the traveling salesman problem. International Journal of

Computational Intelligence Systems, 13(1), 44-55.

Geng, Y., Liu, E., Wang, R., Liu, Y., Rao, W., Feng, S., &

Chen, Y. 2021. Deep reinforcement learning based

dynamic route planning for minimizing travel time. In

2021 IEEE International Conference on

Communications Workshops (ICC Workshops), 1-6.

Helle, J., Poom, A., Willberg, E., & Toivonen, T. 2023. The

Green Paths Route Planning Software for Exposure-

Optimised Active Travel. Journal of Open Research

Software, 11(1).

Huang, F., Xu, J., & Weng, J. 2020. Multi-task travel route

planning with a flexible deep learning framework. IEEE

Transactions on Intelligent Transportation Systems,

22(7), 3907-3918.

Hu, S., Lu, H., Huang, Y., & Xu, K. 2019. Improved Single

Parent Genetic Algorithm for Solving Multiple

Traveling Salesman Problem. Journal of Northeast

Normal University (Natural Science Edition), 51(04),

49-56.

Jiang, H., Bai, K., Liu, H., Li, C., & Fu, Z. 2023. Maximum

problem parallel bounded search. Journal of Computer

Science and Technology, 38(5), 1187-1202.

Li, W., Xia, L., Huang, Y., & Mahmoodi, S. 2022. An ant

colony optimization algorithm with adaptive greedy

strategy to optimize path problems. Journal of Ambient

Intelligence and Humanized Computing, 1-15.

Lin, S., & Kernighan, B. W. 1973. An effective heuristic

algorithm for the traveling-salesman problem.

Operations Research, 21(2), 498-516.

Miao, C., Chen, G., Yan, C., & Wu, Y. 2021. Path planning

optimization of indoor mobile robot based on adaptive

ant colony algorithm. Computers & Industrial

Engineering, 156, 107230.

Sari, I. P., Fahroza, M. F., Mufit, M. I., & Qathrunad, I. F.

2021. Implementation of Dijkstra's Algorithm to

Determine the Shortest Route in a City. Journal of

Computer Science, Information Technology and

Telecommunication Engineering, 2(1), 134-138.

Shivgan, R., & Dong, Z. 2020. Energy-efficient drone

coverage path planning using genetic algorithm. In 2020

IEEE 21st International Conference on High

Performance Switching and Routing (HPSR), 1-6.

Yu, Y., Chen, Y., & Li, T. 2014. Improved Genetic

Algorithm for Solving Traveling Salesman Problem.

Control and Decision, 29(08), 1483-1488.

The Investigation of Advancements in Intelligent Tourism Route Planning Based on Path Generation Algorithms

489