Design and Simulation of an Autonomous Medicine-Serving Robot with Deep Learning-Based Patient Recognition Using ROS

In the modern healthcare scenario, automation plays an inevitable role in reducing human interaction, thereby minimizing the risk of disease transmission and human errors in medication. This paper presents the idea of an autonomous medicine-serving robot designed in Autodesk Fusion 360 and simulated using the Robot Operating System (ROS) and Gazebo. The robot is integrated with a deep learning-based face recognition module to enable autonomous patient identification within a virtual hospital environment. The system is achieved by incorporating a vision node that employs the face_recognition Python library, which utilizes a pre-trained convolutional neural network (CNN) for real-time facial feature extraction and comparison. The robot model, exported in URDF format from Fusion 360, performs mapping and autonomous navigation using LiDAR-based SLAM and Adaptive Monte Carlo Localization (AMCL). A camera node is also integrated with the URDF. Camera data from the simulated environment are processed to recognize registered patients through 128-dimensional facial embeddings, allowing the robot to identify patient details and verify medication details and dosage. The system was successfully executed in ROS Noetic on Ubuntu 20.04, demonstrating accurate detection and recognition in a simulated environment. This work highlights the potential of integrating deep learning perception modules with ROS-based robotic simulations for developing intelligent, contactless healthcare service robots.