Abstract— This paper examines the potential use of robotic manipulators to automate the assembly of burgers at fast food restaurants. More specifically, the paper looks at three basic questions. Is this application feasible with currently available, commercial technology? If not, what is missing? And finally, does it make economic sense to pursue this application? The paper begins by
over viewing the work and the technologies required to automate the work. The paper then details an implementation of machine learning (CNN-based image segmentation) applied to the problem. Next, the paper describes a novel end-effector for handling sliced vegetables that was specifically developed for the burger assembly application. Finally, the paper concludes with an examination of the economic considerations around implementing a cobot burger
chef.
I. INTRODUCTION
The job "flipping burgers" has been around as long as there have been hamburger joints and it is often used as the classic example of a low-skill job. The jobs of cooking and assembling burgers are ubiquitous, however, they have largely resisted automation and are still being performed almost exclusively by human workers.
This paper considers whether now is the time to automate the work of cooking and assembling burgers. Of course, there are different varieties of automation. This paper focuses on the use of robotic manipulators. Thus, the question is whether now is the time to use robotic manipulators to automate this work? This question, in turn, breaks into three more basic questions. Is it feasible with currently available, commercial technology? If not, what is missing? And finally, does it make economic sense to do it?
To understand whether the cobot chef is technically feasible, this paper considers three main subsystems:.
1. Robotic manipulator - Traditional robotic manipulators with the performance specifications (payload, speed, dexterity, etc.) to do the work have been around for decades. The work of assembling burgers is not, however, safe to automate with traditional robotic manipulators because it needs to be performed side-by-side with people. Fortunately, collaborative robotic manipulators (cobots) have been developed to address many of the safety issues with traditional robotic manipulators. Cobots have become a relatively mature technology and are commercially available with performance specifications appropriate for the tasks of cooking and assembling burgers.
2. Image recognition - The cobot burger chef needs to identify and locate objects in an image. For example, it needs to identify a specific portion of sliced vegetable, say a sliced tomato, to pick out of a bin of sliced tomatoes. This task is an obvious fit for AI-based vision systems. Most of us have seen examples of this type of image recognition in the context of autonomous cars, but these cars have extremely powerful (and expensive) computers inside them. Is this task also feasible in the cobot burger chef application using more modest computing hardware? This paper presents a test case implementation.
3. End-effector - People are currently using their hands to do the burger assembly work. Robotic hands have been under development for at least forty years, and they are still many years from being commercially available. [9] If the cobot burger chef is going to be viable in the near term, low-cost end-effectors specialized for the burger application will need to be developed. This paper overviews the development of one such end-effector. This end-effector focuses on the task of picking sliced vegetables from a bin and then placing them onto a burger.
The final section of this paper addresses questions of whether it currently makes economic sense to deploy a cobot burger chef. The potential market size is always a part of economic considerations, and this section presents data around the market size for cobot burger chefs. This section also develops and presents analyses of the Payback Period, Return On Investment (ROI), and Compound Annual Growth Rate (CAGR) of a capital investment in a cobot burger chef. Since the comparison is to human labor, the results of these analyses are tabulated as a function of wage.
download
pdf of complete paper
|
