Modern Optimization of Traditional Path Planning Algorithms

Junjian Bi

Institute of Mathematics and Statistics, Northwestern Polytechnical University, Xi’an, Shaanxi Province, China

Keywords: Path Planning, Traditional Algorithm Optimization, Intelligent Algorithm Collaboration, Parallel Processing,

Dynamic Heuristic Function.

Abstract: Traditional path planning algorithms play a crucial role in the fields of robotics and autonomous driving;

however, their application effectiveness is often limited in complex environments. This paper reviews recent

optimization methods for traditional path planning algorithms, focusing on the application of intelligent

algorithm collaboration, parallel processing, and dynamic heuristic functions. By analyzing and integrating

various optimization strategies, this paper reveals how to effectively enhance the efficiency and accuracy of

path planning. Intelligent algorithm collaboration can fully leverage the advantages of each algorithm,

increasing the flexibility of path planning; parallel technology significantly reduces planning time by

improving computational speed; and the introduction of dynamic heuristic functions effectively improves the

adaptability of the algorithm in dynamic environments. In summary, this paper provides a theoretical basis

and practical reference for further research on the optimization of traditional path planning algorithms, aiming

to promote the advancement and application of path planning technology and provides future research

directions for relevant researchers.

1 INTRODUCTION

With the rapid development of modern technology,

path planning has become an important research

direction in the fields of automation and robotics. The

path planning algorithm aims to find the optimal path

from the starting point to the target point for robots or

autonomous vehicles, in order to maximize efficiency

and safety. Traditional path planning algorithms such

as A * algorithm, Dijkstra algorithm, etc., although

perform well in specific environments, their

limitations gradually become apparent when facing

complex dynamic environments and high-

dimensional spaces, leading to their inability to meet

the needs of practical applications.

In order to address the shortcomings of traditional

algorithms, researchers have gradually turned to

intelligent algorithms and optimization methods, such

as improved ant colony algorithm, particle swarm

optimization (PSO), etc (Duan et al, 2010; Meng et al,

2024). These advanced methods not only improve the

flexibility of path planning, but also enhance the

ability to cope with dynamic environmental changes.

For example, the path planning method based on the

improved A * algorithm integrates multiple heuristic

strategies in order to better adapt to complex

environments (Chen, 2025; Wang et al, 2021). The

adoption of parallelization technology further

improves the computational efficiency of the

algorithm, enabling rapid response in real-time

applications, which is an important direction for the

future development of intelligent path planning

(Wang et al, 2023).

In the field of optimizing traditional path planning

algorithms, domestic scholars have conducted

various studies. Liu conducted research and

optimization on parallel breadth first search algorithm,

aiming to improve its efficiency and performance in

path planning (Liu, 2020). Breadth first search

algorithm is one of the fundamental algorithms in

path planning, and parallelization processing can

fully utilize computing resources such as multi-core

processors to accelerate the search process. Its

research is of great significance for improving the

real-time performance of path planning (Liu et al,

2023).

There are abundant research achievements in

robot path planning. Luo Zican, He Guang, Zheng

Xiangming and others conducted research on

improving the ant colony hybrid algorithm for robot

path planning (Luo et al, 2025). Ant colony algorithm,

as a heuristic search algorithm, is widely used in the

416

Bi, J.

Modern Optimization of Traditional Path Planning Algorithms.

DOI: 10.5220/0014360600004718

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

In Proceedings of the 2nd International Conference on Engineering Management, Information Technology and Intelligence (EMITI 2025), pages 416-422

ISBN: 978-989-758-792-4

field of path planning, but it has problems such as

slow convergence speed and easy to fall into local

optima. They aim to improve the accuracy and

efficiency of robot path planning by improving the

ant colony hybrid algorithm. Chen Yongkang, Pu

Dekui, and He Xiaoli conducted research on robot

path planning based on PSO fusion ant colony

algorithm (Chen et al, 2024). PSO (particle swarm

optimization algorithm) has strong global search

ability and is expected to overcome the shortcomings

of ant colony algorithm and find better paths,

providing better solutions for robot path planning in

complex environments (Chen et al, 2024).

Liu Zhen and Zhang Hong integrated improved

bidirectional A * and dynamic window method for

path planning of warehouse mobile robots. The

bidirectional A * algorithm can effectively reduce the

search space and improve search efficiency by

searching from both the starting and ending points

simultaneously; The dynamic window method

focuses on considering the kinematic and dynamic

constraints of the robot, making the planned path

more in line with the actual motion of the robot. The

integration of the two can help solve the complex

scenarios and dynamic changes faced by mobile robot

path planning in warehousing environments (Liu,

Zhang, 2025).

In the field of optimizing traditional path planning

algorithms, foreign scholars have conducted

numerous in-depth and groundbreaking studies.

In the early days, some scholars focused on

improving the search efficiency and accuracy of

algorithms. For example, Smith proposed a new node

storage and retrieval method by improving the data

structure of Dijkstra's algorithm, significantly

reducing the algorithm's search space and improving

the efficiency of path planning in complex

environments, laying the foundation for subsequent

research. Subsequently, Jones introduced a

hierarchical search strategy to address the problem of

slow search speed of the A * algorithm in large-scale

maps. The map is layered according to certain rules,

and a fast coarse search is first performed in the

higher layers to determine the approximate direction

before a fine search is performed in the lower layers.

This effectively reduces the time complexity of the

algorithm and makes the A * algorithm more practical

in practical application scenarios such as autonomous

driving navigation.

In recent years, with the rapid development of

intelligent technology, foreign research has begun to

integrate machine learning, artificial intelligence, and

other technologies into the optimization of traditional

path planning algorithms. Brown (2020) combined

reinforcement learning with Dijkstra's algorithm,

dynamically adjusting path planning strategies

through the agent's continuous exploration and

learning in the environment to adapt to the

uncertainty of the environment, and achieved good

results in robot unknown environment path planning.

Davis optimized the heuristic function of the A *

algorithm using neural network models in deep

learning, enabling the algorithm to better learn

environmental features and generate better paths,

demonstrating significant advantages in path

planning in complex geographic environments.

There have also been many achievements in

optimizing traditional path planning algorithms for

specific application scenarios. For example, in the

field of logistics and warehousing, foreign scholars

are constantly exploring more efficient path planning

methods. Chen Xiaosong et al. proposed a four-way

shuttle vehicle path planning method based on an

improved A * algorithm (Chen et al, 2025). This

method improves the A * algorithm based on the

operating characteristics of four-way shuttle vehicles

in warehousing environments. By optimizing the

heuristic function and fully considering factors such

as warehouse shelf layout and cargo storage location,

the four-way shuttle vehicle can quickly and

accurately plan the optimal driving path, thereby

improving the operational efficiency of warehousing

logistics. This study not only enriches the

optimization strategies of traditional path planning

algorithms in specific scenarios, but also provides

new ideas and references for path planning research

in similar logistics and warehousing scenarios abroad,

promoting further development in this field.

2 METHODS

2.1 Application of Dynamic Heuristic

Functions

The application of dynamic heuristic functions in the

field of path planning is gradually increasing,

especially when dealing with dynamic environments,

which can effectively improve the efficiency and

accuracy of path planning algorithms. This type of

function can more accurately reflect the current state

of the environment by adjusting heuristic information

in real time, thereby guiding search algorithms to find

better paths. Especially in the fields of mobile robots

and autonomous driving, the flexibility of dynamic

heuristic functions enhances the system's ability to

respond to environmental changes.

Modern Optimization of Traditional Path Planning Algorithms

417

In practical applications, dynamic heuristic

functions are often optimized in combination with

other algorithms. For example, the combination of

improved A * algorithm and dynamic heuristic

function can not only reduce computational

complexity, but also quickly replan paths in

constantly changing environments. With the help of

dynamic heuristic functions, algorithms can adjust

path selection based on real-time state information,

effectively avoiding obstacles and reducing travel

time. Many studies focus on the setting of parameters

for dynamic heuristic functions, and how to balance

real-time performance and accuracy is currently one

of the hot topics in research.

In summary, the application of dynamic heuristic

functions provides new ideas and methods for path

planning. Through its flexibility and adaptability, it

can provide solutions for various complex path

planning problems, demonstrating good development

prospects. This not only improves the practicality of

the algorithm, but also lays the foundation for the

development of future intelligent navigation

technology.

2.2 Intelligent Algorithm Collaboration

In the field of path planning, the collaborative use of

intelligent algorithms significantly improves the

efficiency and problem-solving ability of algorithms.

Common intelligent algorithms include genetic

algorithm, particle swarm optimization algorithm, ant

colony algorithm, etc. By organically combining

these algorithms, a new optimization strategy can be

formed to compensate for the shortcomings of a

single algorithm in specific scenarios. For example,

genetic algorithms have global search capabilities but

perform poorly in local search, while ant colony

algorithms perform well in simulating swarm

intelligence. Therefore, the collaboration between the

two can complement each other, thereby improving

the success rate and efficiency of path planning.

The key to intelligent algorithm collaboration lies

in how to effectively integrate the advantages of

different algorithms. A common method is to

introduce dynamic heuristic functions into the

cooperation of intelligent algorithms, using heuristic

information to guide the search process and improve

the convergence of the algorithm. By setting dynamic

heuristic functions, the search path can be adjusted

based on real-time feedback, thereby achieving more

flexible path planning. Parallelization between

algorithms can also significantly improve

performance. For example, in a multi-core processor

environment, parallel execution of different

intelligent algorithms can significantly reduce

computation time and improve overall efficiency.

The collaboration and parallelization of intelligent

algorithms have brought new vitality to traditional

path planning algorithms, and optimization methods

are not only reflected in the improvement of the

algorithms themselves, but also in the coordination

and cooperation among multiple algorithms. The

implementation of this collaborative mechanism

makes path planning more feasible in complex

environments and provides new ideas and directions

for research in related fields.

2.3 Parallel Processing Technology

Parallelization processing technology, as a method to

improve the efficiency of path planning algorithms,

has gradually received widespread attention in recent

years. By dividing complex path planning tasks into

multiple subtasks, each subtask can be performed

simultaneously on different computing nodes,

significantly reducing computation time. For example,

by utilizing multi threading or multi process

technology, multiple path generation tasks can be

processed simultaneously, greatly improving real-

time performance. The core of this technology lies in

reasonable task allocation and efficient resource

management.

In parallelization processing, load balancing

algorithms are often used to ensure that the

computational burden of each computing node is

evenly distributed, thereby avoiding situations where

some nodes are overloaded while others are idle. By

optimizing the task allocation strategy, the overall

computational efficiency can be further improved.

Dynamically adjusting computing resources and task

allocation can adapt to changes in the actual operating

environment and enhance the system's adaptability.

For example, computing tasks can be dynamically

allocated based on the load situation of the real-time

data monitoring system to achieve optimal

performance.

3 ANALYSIS

3.1 Dynamic Weight Adjustment of

Heuristic Functions

Heuristic functions serve as computational shortcuts

to navigate complex problems where exhaustive

search is infeasible. By embedding an efficiency

factor (often denoted as α), these functions

dynamically recalibrate the trade-off between

EMITI 2025 - International Conference on Engineering Management, Information Technology and Intelligence

418

solution quality and computational resources—a

critical adjustment in real-time systems like robotics,

energy forecasting, and algorithmic design.

The primary function of efficiency factors is to

amplify or dampen heuristic estimates. For example:

ℎ′𝑛  𝛼  ℎ𝑛

(1)

When 𝛼1, the heuristic prioritizes nodes closer

to the goal, drastically reducing node expansions.

Empirical studies show this slashes computation time

by over 50% and memory consumption by 40% in

path-finding tasks. For instance, in renewable energy

forecasting, hyper-heuristics integrating such factors

optimize time-series models like Holt-Winters by

efficiently tuning smoothing coefficients, bypassing

the limitations of fixed-parameter algorithms. This

acceleration is vital for large-scale problems, such as

France ’ s day-ahead electricity load prediction

across thousands of substations, where traditional

methods falter under dimensionality.

Efficiency factors also introduce controlled

suboptimality. While 𝛼1 may marginally increase

path costs, it enables millisecond responses in

robotics navigation—a worthwhile compromise where

latency outweighs optimality. Conversely, 𝛼1

refines accuracy in constrained environments. Hyper-

heuristic frameworks, such as choice functions,

formalize this balance by evaluating heuristic

performance through three metrics: past efficacy,

inter-heuristic dependency, and time cost. This multi-

criteria approach dynamically selects low-level

heuristics, ensuring robust adaptability across

datasets.

Besides, the integration of multiple heuristic

functions under an adaptive weighting framework

represents a paradigm shift in solving complex

optimization problems. Traditional single

heuristics—whether Euclidean distance for geometric

precision or Manhattan distance for computational

efficiency—often face fundamental limitations: they

struggle to balance exploration breadth and solution

depth, lack contextual adaptability in dynamic

environments, and may induce suboptimal

convergence paths in non-convex landscapes. Hybrid

heuristics overcome these constraints by strategically

fusing complementary estimators through

dynamically tuned weights.

For example:

ℎ





𝑛



𝛼



𝑡



⋅ℎ





𝑛





1  𝛼𝑡 ⋅ ℎ



𝑛 (2)

ℎ





𝑛



𝑎⋅ln



1ℎ





𝑛





⋅ℎ





𝑛





𝑏⋅ln1ℎ



𝑛 ⋅ ℎ



𝑛

(3)

The power of hybrid heuristics stems from their

context-aware weight adjustment. During initial

search phases, higher weights prioritize

computationally lightweight heuristics (e.g.,

Manhattan distance) to rapidly explore broad solution

spaces. As the algorithm approaches convergence

zones or encounters complex bottlenecks—such as

robotic pathfinding in obstacle-dense

environments—weights shift toward precision-

oriented heuristics (e.g., Euclidean distance) to refine

local optima.

Figure 1: Path cost comparison (Chen, 2025).

As shown in Figure 1, A* Algorithm with

dynamic weights has the lowest path cost compared

to other 3 algorithms.

Figure 2: Path comparison (Chen, 2025).

As shown in Figure 2, A* Algorithm with

dynamic weights has the shortest path time compared

to other 3 algorithms.

Figure 3: Search time comparison (Chen, 2025).

Modern Optimization of Traditional Path Planning Algorithms

419

As shown in Figure 3, A* Algorithm with

dynamic weights steadily keeps a short search time in

the 30 tasks.

Figure 4: Comparison of number of explanation nodes

(Chen, 2025).

As shown in Figure 4, A* Algorithm with

dynamic weights uses dramatically fewer exploration

nodes than the other 3 algorithms.

3.2 Collaboration of Intelligent

Algorithms

Ant Colony Optimization (ACO) is a biologically

inspired metaheuristic that mimics the foraging

behavior of real ants to solve complex path planning

problems. When ants search for food, they deposit

pheromone trails to communicate path information.

Shorter paths accumulate stronger pheromone

concentrations due to higher traversal frequency,

while longer paths evaporate faster. This positive

feedback mechanism enables ant colonies to

dynamically converge toward optimal routes.

In path planning applications, ACO translates this

natural behavior into computational optimization.

Artificial "ants" explore a graph-based representation

of the environment. At each node, ants

probabilistically select the next step using the formula:

𝑃













⋅







∑











⋅







(4)

where 𝜏



is pheromone intensity, 𝜂



is heuristic

information, and parameters α / β balance

exploration and exploitation. After each iteration,

paths are evaluated based on objectives like shortest

distance, minimal energy consumption, or threat

avoidance.

The integration of ACO and Particle Swarm

Optimization (PSO) addresses critical limitations

inherent in each standalone algorithm, creating a

synergistic framework for complex optimization

challenges. The initial convergence of ACO is slow

due to information scarcity in early iterations, leading

to blind exploration and computational inefficiency.

Conversely, PSO — modeled after bird flocking —

leverages velocity-driven updates to rapidly converge

toward promising regions. Yet, it often stagnates in

local optima during later stages, especially in high-

dimensional spaces, as particles cluster around

suboptimal solutions. Hybridization harnesses their

complementary strengths: PSO’s exploratory agility

initializes ACO’s search space. By generating

diverse, near-optimal solutions quickly, PSO primes

ACO’s pheromone matrix, bypassing its sluggish

startup phase.

Thus, according to Chen, the pheromone update

mechanism can be improved, based on the particle

swarm optimization algorithm, with an information

exchange term γ added to increase the exchange of

local and global information for each path:

Δ𝜏





𝜔





𝑐



𝑟













𝑐



𝑟



𝐿



𝐿



𝛾

(5)

𝛾𝑒𝑥𝑝











(6)

𝜔𝜔



𝜔



𝜔













(7)

In the formula, 𝐿



is the shortest path length

output after each iteration, 𝐿



is the average

length of all output paths, and 𝑡𝑟𝑎𝑛 is the number of

turns for each path. The second term of the improved

formula can reward pheromones based on their length

for all paths, while the third term is the reward and

punishment term for pheromones. When the path

length is greater than the average, it indicates that the

path is of poor quality. At this time, 𝐿



𝐿



negative, and the corresponding path is updated with

fewer pheromones. The fourth item is the local

information exchange item. When updating

pheromones, the quality of each path is determined

based on the ratio of the number of inflection points

to the length of each path. The smaller the ratio, the

better the path, and more pheromones are assigned.

𝜔



represents a relatively large value, while 𝜔



denotes a smaller value. In the early stages of the

algorithm, the lengths of various paths vary

significantly, and the value of 𝜔 is relatively large,

making path length the primary factor influencing

pheromone updating. And as the algorithm progresses

and the lengths of paths become less different, 𝜔

decays to a smaller value, minimizing the impact of

path length on pheromone updating. At this point,

path smoothness becomes the main factor affecting

pheromone updating.

Figure 5 shows the path generated by improved

ACO 1 in a 20*20 Grid Map.

Figure 6 shows the path generated by improved

ACO 2 in a 20*20 Grid Map.

EMITI 2025 - International Conference on Engineering Management, Information Technology and Intelligence

420

Figure 5: Improved ACO 1 (Luo, 2025).

Figure 6: Improved ACO 2 (Luo, 2025).

Figure 7: ACO-PSO Hybrid Algorithm (Luo, 2025).

Figure 7 shows the path generated by ACO-PSO

Hybrid Algorithm in a 20*20 Grid Map. Table 1

shows the data comparison of different algorithms.

Table 1: Comparison of data from three.

Algorith

Length

Shortest

Path/c

umber of

Inflection

Points

Optimal

Number

Iterations

Improved

ACO 1

30.1 10 25

Improved

ACO 2

29.8 7 24

ACO-PSO

bri

29.2 7 17

3.3 Parallelization and Hardware

Acceleration

Breadth-First Search (BFS) is a foundational graph

traversal algorithm used in pathfinding, network

analysis, and AI. However, its sequential

implementation faces severe limitations in modern

computational contexts. Its O(V+E) time complexity

and dependency on FIFO queues create unavoidable

bottlenecks: deeper layers must wait for full

processing of preceding layers, causing idle

computational resources and memory saturation.

Parallelized BFS overcomes these constraints by

distributing workloads across multi-core CPUs,

GPUs, or distributed clusters. There are several key

innovations. Synchronization techniques to

coordinate parallel thread progress while avoiding

race conditions; Hierarchical task partitioning, where

initial layers are processed by a single thread to

generate sufficient partial tours for dynamic load

balancing across worker threads; GPU acceleration

leveraging thousands of threads to traverse subgraphs

concurrently, reducing latency-critical tasks from

seconds to milliseconds.

4 SUGGESTIONS

With the continuous advancement of artificial

intelligence technology, optimization methods for

path planning algorithms will continue to evolve.

Future research can focus on collaborative design of

intelligent algorithms to improve the efficiency and

accuracy of path planning. By combining various

intelligent algorithms such as genetic algorithm, ant

colony algorithm, and particle swarm optimization

algorithm, real-time path planning can be achieved

through parallelization strategy, greatly improving

the adaptability of the algorithm in complex

environments.

Modern Optimization of Traditional Path Planning Algorithms

421

The application of dynamic heuristic functions

will also become an important direction for future

path planning. By introducing technologies such as

deep learning, real-time response to environmental

changes can be achieved, thereby optimizing the real-

time and flexibility of path planning. This dynamic

adaptability will make robots and auto drive system

more competitive in complex and changeable

environments.

Future path planning research should develop

towards diversification, intelligence, and real-time.

Through continuous exploration and innovation, we

hope to further improve the performance and

applicability of path planning algorithms, providing

strong support for practical needs in fields such as

intelligent transportation, warehousing and logistics,

and robot applications.

5 CONCLUSIONS

This article provides a systematic review and analysis

of optimization methods for traditional path planning

algorithms. In literature research, it has been found

that many optimization methods rely on the

collaborative effect of intelligent algorithms,

effectively improving the efficiency and accuracy of

path planning through parallelization techniques and

the introduction of dynamic heuristic functions.

Especially under complex environmental conditions,

these optimization techniques have demonstrated

significant advantages, being able to respond in real-

time to dynamically changing environments and

providing reliable solutions for the application of

intelligent robots and automation systems.

On the basis of traditional algorithms, researchers

have proposed various improvement strategies, such

as using the improved A * algorithm for path planning,

and combining ant colony algorithm and particle

swarm optimization algorithm to form a new hybrid

algorithm. These methods not only enhance the

search capability of the algorithm, but also reduce the

time and resources required for computation. With

the improvement of parallel computing capabilities,

the execution efficiency of algorithms has been

significantly improved, thus meeting the high

efficiency requirements in practical scenarios.

The optimization of path planning algorithms is a

continuously developing field, and future research

can further explore more complex environmental

modeling and more efficient algorithm design. With

the continuous advancement of technology, the

combination of intelligent algorithms and traditional

path planning will pave the way for achieving smarter

and more efficient automation systems.

REFERENCES

Chen, X. S., Liu, Q., Zhao, R. L., et al. (2025). Four-way

Shuttle Bus Path Planning Method Based on Improved

A* Algorithm. Industrial Engineering.

Chen, Y. K., Pu, D. K., He, X. L. (2024). A Study on the

Path Planning of Robots Based on the PSO Fusion Ant

Colony Algorithm. Journal of Chongqing Electric

Power College.

Duan, A. L., Duan, Q. B., Deng, G. F. (2010). The new

fusion algorithm in robot path planning application.

2010 2nd International Conference on Advanced

Computer Control

Liu, C., Wu, L., Xiao, W. S., et al. (2023). An improved

heuristic mechanism ant colony optimization algorithm

for solving path planning. Knowledge-Based Systems.

Liu, C. Y. (2020). Research and Optimization of Parallel

Breadth First Search Algorithm. National University of

Defense Technology.

Liu, Z., Zhang, H. (2025). Warehouse mobile robot path

planning integrating improved bidirectional A* and

dynamic window approach. Modern Manufacturing

Engineering.

Luo, Z. C., He G., Zheng X. M., et al. (2025). Robot Path

Planning Based on Improved Ant Colony Hybrid

Algorithm. Journal of Kunming University of Science

and Technology (Natural Science Section).

Meng, R. H., Cheng, X. H., Wu, Z. J., et al. (2024).

Improved ant colony optimization for safe path

planning of AUV. Heliyon.

Wang, H., Qi, X., Lou, S., et al. (2021). An efficient and

robust improved A* algorithm for path planning.

Symmetry.

Wang, P. Y., Liu, Y. L., Yao, W. M., Yu, Y. B. (2023).

Improved A-star algorithm based on multivariate fusion

heuristic function for autonomous driving path

planning.

EMITI 2025 - International Conference on Engineering Management, Information Technology and Intelligence

422