Design And Research Of Road Condition Information System Of Mobile Platform In Semi-enclosed Area

With the substantial improvement and large-scale promotion of intelligent technology,automated driving equipment has been increasingly intensively used in different fields,ranging from household sweeping robots to industrial mobile robots,b becoming more and more common in all walks of life.In the face of currently intense global epidemic and normalized epidemic prevention and control in China,automatic driving equipment that does not require direct human contact has become an important way to achieve the goal of “zero contact” in the“last mile” and controlled communities in the logistics process of various materials and minimize risks potentially caused by contact.Accurate road recognition and obstacle avoidance are the key to effective unmanned autonomous driving equipment.Therefore,the reliability of its road condition information detection system is particularly important.This thesis mainly intends to design and study road condition information system for medium-and low-speed mobile platform in the semi-enclosed area.Semi-enclosed area refers to an area where non-motor vehicles and pedestrians are the main traffic,but motor vehicles are prohibited from passing through and are characterized by the slow movement of objects.Based on the above research purposes and previous research on the status quo and development trend of unmanned platforms at home and abroad,the author takes the image recognition direction of unmanned platforms as the research focus of this paper.An embedded edge computing platform is formed by installing embedded edge computing module and various sensors on a crawler trolley as a mobile mechanism.This paper proposes a solution of road condition information system on mobile platform,which solves the problem of information acquisition ability of embedded edge computing platform.In this paper,nine kinds of common obstacles are selected as the identification targets of the mobile platform road condition information system to analyze the image features of the obstacle targets.The deep learning YOLOX model with excellent detection effect and strong expansion ability is used to detect road conditions and identify common obstacles,such as people,dogs,cats,etc..Aiming at particular low-speed road condition information targets,the backbone network of the YOLOX algorithm is improved by adding attention mechanism to improve the recognition performance of obstacle targets.The improved algorithm program based on YOLOX is deployed on the Jetson Nano embedded edge computing platform.On the platform,image recognition module,ultrasonic module and GPS positioning module are used to identify target types,measure relative distances,and extract target location information,which are finally recorded by the platform,and store relevant information.Road condition information summarized on the embedded edge computing platform will provide data support for subsequent route optimization,thus laying a solid foundation for the safe operation of the platform.