Building a Navigation System for a Shopping Assistant Robot from Off-the-Shelf Components

Abstract

A primary goal of developing robots is to relieve people from hard work or help them in difficult situations. An example for such a situation is a shopping assistant robot that supports people who need help in their daily life. Such a mobile robot has to be able to deal with dynamic environments (moving people) and navigate safely and efficiently. In the present paper, we provide a system description of our navigation strategy for an assistant robot that is developed for autonomous operation in a supermarket. Creating an appropriate experimental environment can be quite challenging, and access to the application place can be limited (e.g., a real supermarket). Therefore we complement our real robot with a digital twin and describe our approach to create a suitable simulation of a real-world supermarket. Further, we discuss how off-the-shelf software can be used to implement a three-stage navigation strategy (planing a route with a TSP solver, global path planning, and incorporating dynamic obstacles in an optimization-based TEB approach) that is suitable for environments with dynamic obstacles. The paper presents our approaches to create a 3D map from 2D floor plans, as well as the preprocessing of the sensor data for usage in the TEB planner. Finally, we provide our hands-on experience with implementing a complex state-machine using the graphical RAFCON framework.