Research On Efficient Medicine Identification And Autonomous Distribution Technology For Mobile Manipulation Robot In Ward Scenarios

With the development of robotics and artificial intelligence technology,various kinds of robots have gradually entered into our daily life,which have liberated people from simple and trivial work.In recent years,with the emergence of the COVID-19,the shortage of medical resources has become increasingly obvious.However,hospital wards still require a large number of medical staff for the distribution of patient medicines.In order to address this problem,this paper focuses on the medicine identification and autonomous distribution technology for mobile manipulation robots in the ward environment.The main research contents are summarized as follows:(1)In order to quickly identify the medicine box,this paper proposes a lightweight feature extraction network to improve the detection efficiency of the Faster R-CNN algorithm.In this work,we introduce the transfer learning to improve the timeliness of model detection and realize the rapid identification of medicine boxes on the premise of ensuring the accuracy.In addition,this paper propose to utilize the OCR algorithm to identify the text on the surface of the medicine,and realizes the rapid recognition of the text on the medicine box.(2)In order to determine the pose of the medicine box to assist the robot grasping,this paper adopts the Iterative Closest Point(ICP)algorithm to realize the rapid registration of the point cloud of the pill box and determine the pose of the medicine box.The grasping pose of mobile manipulation robot is then determined accordingly.Quantitative experiments have also been designed to verify the accuracy and timeliness of the registration algorithm.(3)In order to improve the distribution efficiency of robot in dynamic humanrobot coexisting ward environment,we propose a discomfortable function based on pedestrian psychology.We incorporate the discomfortable function into the pedestrian model so as to enable the robot to predict the movement of pedestrians.The experimental results reveal that the proposed method significantly improves the success rate of autonomous navigation in pedestrian interference environments.(4)Furthermore,we also build the software and hardware system of the mobile manipulation robot platform and design the experimental environment.The medicine delivery experimental studies have been carried out under pedestrian interference.The experimental results show that the robot in the ward scene can efficiently identify the pill box and complete the medicine delivery autonomously.In summary,this paper proposes a feature extraction network model to achieve rapid detection of pill boxes;by incorporating the discomfortable function into the RiskRRT algorithm,the robot can successfully navigate in the human-robot coexisting ward environment.We have integrated the proposed algorithms into the mobile manipulation robot and realize the rapid identification and grasping of the medicine box and efficient autonomous navigation in the ward environment.