---

Introduction

The Traveling Salesman Problem (TSP) presents a classic conundrum in computational theory: given a set of cities and the distances between each pair, how can one find the shortest possible route that visits each city once and returns to the starting point? While the problem statement itself is straightforward, devising efficient solutions and testing them requires extensive experimentation with diverse datasets.

Historically, the most common format for storing TSP datasets has been the TSPLIB format, which generates a separate file for each sample. This method, while universally adopted, is far from storage-efficient, leading to significant overhead when managing large volumes of data. Each file needs to be read, parsed, and loaded separately, which can be cumbersome and time-consuming, especially for large-scale studies involving thousands of TSP instances.

Recognizing the limitations of the TSPLIB format and the absence of an optimized, well-known method in the literature for efficient TSP dataset storage, we developed the TSP Dataset Generator. This innovative tool leverages the versatility and accessibility of JSON files to streamline the storage and retrieval of TSP data. By encapsulating multiple TSP instances within single JSON files, our approach not only enhances storage efficiency but also simplifies data handling, making it easier for researchers and developers to focus on solving the problem rather than managing the data.

The Need for Custom TSP Datasets

The Traveling Salesman Problem requires diverse and comprehensive datasets for effective algorithm development and testing. Traditional datasets often fall short in providing the variability and specificity needed to robustly test and benchmark new TSP solutions. This is where the need for custom TSP datasets becomes evident—datasets that not only vary in size but also in the complexity and configuration of the problems they present.

Flexibility in Dataset Size

One of the significant advantages of the TSP Dataset Generator is its ability to produce datasets with any number of cities. This flexibility is crucial for researchers and developers who need to test algorithms at different scales, from small instances that might be solvable with exact methods to larger instances that require heuristic or approximate approaches. By allowing users to specify the number of cities, the generator makes it possible to tailor datasets to the specific needs of the research or the computational limits of the algorithms being tested.

Customizable Grid Dimensions and Step Size

Another critical feature of the generator is the ability to modify the grid dimensions within which the cities are placed. This adjustability enables the creation of datasets that can simulate various geographical constraints and scenarios.

Furthermore, the inclusion of a configurable step size adds an additional layer of customization. By setting the step size, users can define the granularity of the grid:

• Smaller Step Size: A smaller step size leads to a finer grid, which allows for cities to be placed more closely together. This setting enables the creation of datasets where cities are more densely packed, providing a higher level of detail and more complex routing challenges.
• Larger Step Size: A larger step size increases the spacing between potential city locations, making the problem more discrete and reducing the complexity of city placement. Setting the step size to a very large value allows for cities to be placed more sporadically, simulating scenarios with less frequent but strategically important locations, which reflects certain real-world logistical challenges more accurately.

These adjustments to the step size directly impact the difficulty and complexity of the TSP instances generated, offering a spectrum from highly detailed, dense configurations to more strategic, sparsely populated maps. This flexibility is essential for tailoring datasets to the specific challenges and hypotheses being tested in TSP research.

Addressing the Gap in Dataset Variability

These features address a critical gap in TSP research tools—a lack of dataset variability and configurability. Most available TSP datasets do not offer the flexibility needed to thoroughly test different aspects of TSP algorithms, such as their scalability, efficiency, and accuracy across various types of geographical distributions and problem sizes. The TSP Dataset Generator not only fills this gap but also empowers users to explore new dimensions in algorithm development and testing.

Accessing the Generated TSP Datasets

Throughout the development of the TSP Dataset Generator, we have created a variety of datasets tailored to different research needs and algorithm testing scenarios. These datasets are a testament to the flexibility and utility of the generator, showcasing its ability to produce highly customizable TSP problems. Below, we highlight several key datasets that have been generated, each serving a unique purpose in the realm of TSP research.

Datasets:

1. tsp_10_50k_u_100x100

• Characteristics: Consists of 10 cities per problem, with cities distributed within a 100x100 grid. Ideal for testing basic TSP solvers and introductory algorithmic training.
• Link

2. tsp_10_50k_u_100x100

• Characteristics: Consists of 10 cities per problem, with cities distributed within a 100x100 grid. Ideal for testing basic TSP solvers and introductory algorithmic training.
• Link

Create Your Custom TSP Dataset

For researchers, developers, and enthusiasts looking to tailor TSP datasets to their specific needs, our GitHub repository offers all the necessary tools and instructions. The TSP Dataset Generator hosted on GitHub provides a comprehensive, user-friendly platform for creating customized TSP datasets with variable city counts, grid dimensions, and other parameters.

Explore Our GitHub Repository
Repository Link: TSP Data Generator
Features:

• Complete source code for the TSP Dataset Generator.
• Detailed documentation on how to configure and run the generator.
• Examples of configuration files for creating diverse TSP datasets.

Conclusion

The TSP Dataset Generator offers an innovative solution for creating diverse and customized datasets, addressing the specific needs of the TSP research community. By providing tools that allow for easy customization and generation of datasets, we hope to foster further research and development in solving the Traveling Salesman Problem. Explore our Github repository to start creating your tailored datasets and contribute to advancing this field of study.