Loan Bernat

Introduction

During my internship at Ostfalia (Wolfenbüttel, Germany) - University Of Applied Sciences, I had to program an application for Sawyer robot to transform it into a bartender. It had to work with robotics waiters (Pepper robot) and perform some of classical actions of a bartender: recognize an order, opening bottles, serving cocktails in glasses

Application Basis

To use Node-Red to command the Sawyer, I needed to create 3 docker containers : the first one for the node-red part, the second one is for the ROS environment and scripts, the last one is a rest-server to make the communication between the two first. More information in my technical note / report (link comming soon)

Fixed Position V1.5

This version assumes that plastic cups and bottles are always placed in the same place every time. Results are satisfaying. An order is prepared in 1 minute with 100% accuracity (taking/pouring).

Object Detection with Yolov5 V2.0

I used Yolov5 to train models with my own datassets labelled with RobotFlow. I integrated this model directly in my ROS scripts to allow the robot to adapt its trajectory in real time. You can find this on my github repository