Research on improved A-star algorithm for global path planning of unmanned logistics vehicles

Abstract

Global path planning process of unmanned logistics vehicles, such as difficulty in improving accuracy, long operation time, and poor driving stability, this paper improves map construction, acquisition node mode, and expansion node mode. Firstly, road nodes instead of obstacle nodes are used in map construction so that the search times and search time are reduced. Secondly, UTM nodes are collected in the field to construct scatter maps instead of grid maps, which solves the problem of long searching time under high precision caused by the positive correlation between grid accuracy and path accuracy in the traditional A-star algorithm. Finally, the k- nearest neighbor is used to expand the sub-nodes instead of the nine-grid method to improve the path accuracy of the Astar algorithm, reduce the search time, and improve the driving stability of unmanned logistics vehicles. Based on the unmanned logistics vehicle experimental platform of an intelligent vehicle research and development company in Tianjin, this paper uses the proposed algorithm to carry out global path planning in a specified area of 27,800 square meters. The experimental results show that the centimeter- level (10 significant digits of the path node) searches were reduced to 765. Compared with the A-Star algorithm under the grid map, it is reduced by 99.88% and the search time was reduced from 41884s to less than 2s. At the same time, by changing the value of k, several controlled experiments were conducted to compare the number and distance of nodes in each group. Finally, when k is 5, the number of nodes is 117, and the distance between nodes is 0.5 to 2m, the unmanned logistics vehicle can run continuously and stably.