To address the low efficiency and easy diagonal crossing of obstacles in complex building environments when using the D

*

algorithm

an improved D

*

path planning algorithm is proposed.First

an undirected graph of key nodes is generated through Delaunay triangulation and node optimization

which replaces the traditional grid map and improves search efficiency.Second

a virtual node connecting all exits is introduced to optimize the search strategy and further improve execution efficiency.Finally

obstacles and study-area buffers are established to improve path planning accuracy and effectively avoid paths from crossing obstacles.Validations on virtual envi

ronments and Stanford teaching building data show that the optimized algorithm reduces computation time by 77.56% and significantly reduces diagonal obstacle crossings.Especially in environments with large areas and multiple exits

both the search efficiency and path accuracy are significantly improved.