The traditional Dynamic Window Approach (DWA) with constant weight values of the evaluation function leads to the inability of obstacle avoidance for the Automated Guided Vehicles (AGV) to perform obstacle avoidance and path planning in the complex environment. Effective avoidance of complex obstacles requires adaptive weight adjustment to address the evaluation function's challenges. This paper proposes an adaptive DWA (ADWA), which introduces neural network training on the basis of the Mamdani DWA (MDWA). Firstly, the Mamdani type fuzzy controller is designed, and then the adaptive neuro-fuzzy controller is obtained by neural network training. Then, experiments are carried out through the MATLAB simulation environment. The simulation experiment results show that the improved DWA compared to traditional DWA can make the AGV pass the obstacle environment with a better trajectory and reduce the time. The improved DWA improves the autonomous obstacle avoidance capability of AGVs, which not only perfectly fits our task requirements, but also has apparent scientific and practical significance in developing AGV autonomous obstacle avoidance technology.INDEX TERMS fuzzy control, dynamic window approach, neural network, automatic guided vehicle

Shelter identification is the fundamental issue to make shelter-transporting automated guided vehicle (AGV) effectively detect and transport shelter. Actively identifying shelter has an important problem of high accuracy but slow speed for a complex model, and fast speed but low accuracy for a simple model. However, all kinds of target detection algorithms available have low detection accuracy and speed. In this paper, the model YOLOv5n6* is developed based on the modified YOLOv5 model by selecting different model structures, introducing an attention mechanism, and improving loss function and non-maximum suppression (NMS). Then, the experiments for shelter recognition were carried out using the model YOLOv5n6*. The experimental results show that the box_loss is reduced by 1.2%, the mAP_0.5:0.95 is improved by 2%, and the detection accuracy is improved by 0.87% for the improved model YOLOv5n6* compared with the YOLOv5n6. However, the YOLOv5n6* size is only 7.2M, and the detection time is increased by 0.2ms. So it is proved that the modified model YOLOv5n6* not only has a significant improvement in the shelter detection ability but also has strong robustness, which meets both the requirements of the recognition accuracy and the detection speed.INDEX TERMS YOLOv5, Automated guided vehicle, Target detection, Attention mechanism I. INTRODUCTIONField shelter hospitals are of great significance to accomplish medical treatment for the patients in medical support, emergency prevention and control of infectious diseases, and emergency medical rescue of sudden disasters. The field shelter hospital consists of multiple shelter modules, and its deployment involves the cooperation of multiple modules, so the shelter transfer has become an important step during its deployment process. The shelter should be transported to the assignment position to satisfy docking requirements among shelters. Shelter-transporting efficiency seriously affects the deployment of shelter hospitals [1~6]. Shelter-transporting automated guided vehicle (AGV) as an intelligent transportation device, has higher intelligence and can complete the shelter-transporting at a faster speed and higher precision. However, the shelter-transporting AGV faces extremely serious problems that shelter-transporting AGV identify shelters have low detection accuracy and slow speed. Shelter-transporting AGV combining computer vision can effectively solve the above problems during the deployment of field shelter hospitals. The target detection algorithm can accomplish the shelter accurate recognition, then improves shelter-transporting AGV efficiency and reduces the deployment time of field shelter hospitals.Recently, many scholars have proposed different target detection models [6~8]. For example, the two-stage algorithm model is represented by region-network backbone network series [9], and the two-stage target detection network use region extraction operation. Firstly, the

No abstract