Design And Implementation Of A Caddle Robot Prototype With Automatic Following Funcation

With the rapid development of computer-related technology and sensor technology,sensor technology and algorithms are gradually applied to the control of robots,which has spawned a large number of industrial robots used in industrial production environments.However,there are relatively limited research on human-computer interaction mobile robots.Based on UWB sensor positioning system and computer vision detection and tracking algorithm,this study aimed to design and implement a caddie robot prototype with automatic following function.This paper focuses on the automatic following algorithm used in caddie robot.By improving the UWB positioning scheme commonly used in indoor environments,the caddie robot automatically follows human positioning targets and can be applied to provide users with good services in environments such as golf courses.When a UWB sensor uses a single label with multiple base stations,the phase difference generated by the time sequence collected data of multiple base stations will affect the accuracy of the positioning results.This study provided a consumption method that weakens the concept of data usage based on time sequence.It linearizes the discrete data through the local linearization algorithm which completely decouples the data collection and use.There is no need to pay attention to the data collection time during use,thereby solving the phase difference problem.Compared with the direct use of discrete data,the positioning result after linearization increases by about 1.2 cm in distance and about 2.8 degrees in angle.The problem caused by the installation of the UWB base station on the robot is that the base stations are very close,which causes the positioning results to be very sensitive to errors.In particular,the angle calculation results between the robot and the human are very susceptible to errors,and the fluctuations can even reach 15° to 20°.This study solves the problem of angle by combining the visual detection and tracking algorithm.The detection and tracking algorithm designed in this study only tracks the target of interest.The unique target of interest is determined by combining the IOU matching mechanism and the UWB angle positioning result.By combining the UWB positioning system and the detection and tracking algorithm,the distance error of the positioning output is about 10cm,and the angle fluctuation range is about 2°.Compared to a UWB output alone,the angle is significantly improved.The vision algorithm used on the robot needs to have real-time performance.This study uses heterogeneous computing to accelerate the vision-related algorithm,so that the vision algorithm runs on the GPU to ensure the fast response ability of the algorithm.After GPU acceleration,the performance of the algorithm is increased by about 7 times,and the processing time of a single image on the TX2 is about 20ms,so that the algorithm can be applied to the caddie robot.In conclusion,based on the above research results,the caddie robot prototype is designed according to the actual application scene of caddie robot.Aiming at the soft ground environment of golf course and the actual environment of small slope ground,the automatic following caddie robot prototype is designed and realized.