Session: 02-01-01: Product and Process Design

Paper Number: 144586

144586 - A Human-Centric Architecture for Assistive Cobot Workflow 

At first, the robots (or traditional industrial manipulators) were designed to execute predictable, repetitive and orderly tasks  for use in a manufacturing line. The speed, precision and high-repeatability of an industrial robot often makes them economically viable for high-volume production. For safety reasons industrial robots are caged and therefore unable to share the same workspace with humans. Cobots, on the other hand, with an ethos of sharing a common workspace with humans is crucial in a production environment as humans generally excel in tasks that lean towards dexterity, cognitive reasoning and decision making. These mutually-exclusive advantages paired with speed and precision of a robot in a shared workspace has potential to carry out tasks which are forsaken by industrial robots. 

 

There are several classes of workflow where a multitude of robots (including mobile robots, and manipulators) benefit from working alongside humans for an efficient and optimized operation with a motive of repurposing the workers to a higher skilled job while tasking the robots to take over all possible assistive tasks. In our case study, we have devised a mock workflow where a worker is performing a dexterous task of assembling an object that consists of several aluminum extrusions, fastening a few electronics components and validating the design all the while interacting with the robotic-system using voice and gestures. During this workflow all the components needed are fetched and hand-delivered to the worker. The robot support system consists of two cobots on a mobile robot platform capable of navigating across the floor. 

 

Through the case study, our contribution is the approach in which we envision the workflow which is designed in layers. We put the worker at the center of the followed by the objects in question, and then the robot support system. In this Human Robot Interaction (HRI) approach we have decoupled the robot support system, and made it adaptive to the high-level motion commands. We believe that the two independent systems now can be designed separately where a given workflow-model can be paired with any robot-support system with minimal modifications. This type of design thinking also helps in deploying robots in a domestic environment where the robotic-system is primarily operating on the task in hand and adapting themselves based on the environment. The semantics of low-level robot commands are handled by machine learning.

 

Currently we are working on having multiple participants try the aforementioned mock workflow all by themselves and then with the robot-support system with a goal to evaluate the efficacy of this system in the sense of Task Load Index. We will present this new HRI approach and publish the findings at the conference.

 

Presenting Author: Yohmei Harada Megachips LSI USA

Presenting Author Biography: Yohmei Harada is an engineer at Megachips LSI.

Authors:

Yohmei Harada Megachips LSI USA
Manoj Sharma Santa Clara University
Christopher Kitts Santa Clara Univeristy