This paper presents a dataset of grid-based maze environments constructed from orthogonal polygonal structures. The orthogonal polygons are generated using a constructive generation framework. Building upon these polygonal structures, the present work further constructs maze environments by transforming polygons into grid-based representations, automatically generating start–goal configurations, and producing associated benchmarking data. The dataset covers a wide range of geometric complexities by varying the number of polygon vertices. Each maze instance is represented on a discrete grid, where polygon boundaries are treated as obstacles and interior regions as navigable space. The data are provided in a structured format, including orthogonal polygon representations, grid-based mazes, visualization images, and metadata, allowing direct use in computational experiments without additional preprocessing. Benchmark results are included for several classical path planning algorithms, including Breadth-First Search (BFS), Dijkstra’s algorithm, A*, Probabilistic Roadmaps (PRM), Rapidly-exploring Random Trees (RRT-Connect), and RRT*. For each instance, performance indicators such as success rate, path length, and execution time are reported to illustrate basic usage of the dataset. In addition, an extended stochastic evaluation is provided, where sampling-based methods are executed over five independent runs and RRT is further analyzed under a convergence-based stopping criterion, offering a more comprehensive view of variability and solution refinement. The dataset can support benchmarking and comparative evaluation of path planning algorithms, as well as studies on the influence of geometric structure on navigation tasks. In addition, it may be used in related application contexts such as procedural maze generation in games, simulation environments with controlled spatial complexity, and as a source of data for learning-based approaches to navigation, where both maze structures and algorithm-generated trajectories are available. The data generation pipeline and associated resources are provided to facilitate reproducibility and further reuse.
Nguyen Kieu Linh (Fri,) studied this question.