Research Introduction
Regional exploration has always faced many challenges in various applications. Although significant progress has been made in small-scale exploration tasks, there are still some problems in large-scale scene exploration, especially due to insufficient viewpoint quality and insufficient attention to boundary areas, which have caused many difficulties. Specifically, the problems faced by large-scale exploration include excessively long flight paths and inefficient utilization of boundary information. In response to these issues, this article proposes an optimization method based on regional exploration and global loss function, aiming to effectively utilize boundary information. The specific method is as follows: Firstly, we determine safe areas near the environmental boundaries and generate Gaussian distribution viewpoints within these areas. Then, the exploration area is divided into multiple sub areas, and the priority of each area is calculated based on the traversal order to determine the key areas for exploration. In addition, we balance the weights between different exploration areas through a boundary gain mechanism, further reducing the phenomenon of drone backtracking.
Results
Our experimental results indicate that our method has achieved significant performance improvements on multiple benchmark datasets.
Video Display
The following is a demonstration video of our method, demonstrating how the algorithm can be applied in real-time scenarios.
Download the paper(Pending )
Click the following button to download the PDF version of the paper：
Download PDF