Multi-AUV collaborative search for multiple targets in a dynamic obstacle environment
Yujuan Song, Tailai Pang, Daqi Zhu · 2021 China Automation Congress (CAC) · 2021
Aiming at the problem of dynamic obstacles encountered in the path of multi-AUV collaborative search for multiple targets under the two-dimensional grid map, a multi-AUV search algorithm based Glasius Bio-Inspired Neural Network (GBNN) is proposed. First, a two-dimensional grid map based on the AUV search environment is built, and a corresponding two-dimensional biologically inspired neural network model based on the two-dimensional grid map is realized. Each neuron in the neural network corresponds to each location unit of the grid maps one by one; Then, dynamic obstacles are introduced into the search path, and the AUV path is independently planned according to the activity output value of GBNN. Finally, the experiments show that the target search algorithm proposed in this article can not only avoid static obstacles in the raster map, but also avoid dynamic obstacles to search for multiple targets. It compares with the traditional biologically inspired neural network algorithm (BNN). In contrast, the algorithm has a shorter search path, shorter time-consuming, and stronger real-time performance.