Two fingers suffice

In just a few hours a robot can be fitted with new fingers and reprogrammed for a new assembly task. For his doctoral project, Mohammadali Honarpardaz has developed a method to construct rapidly a new configuration of robot fingers designed for a new task.

A robot hand meet a human handA robot hand just need two fingers. Photo credit: iStock photo"From a purely mathematical point of view, more than two fingers are never required for the different tasks that a robot will carry out. Research into anthropomorphic five-fingered hands for robots is for home service robots that are to help old or sick people. We want such robots to be similar to humans as both are going to share same environment," says Mohammadali Honarpardaz, who has recently defended his doctoral thesis in the Division of Machine Design at Linköping University.

Horizon 2020 programme

Mohammadali Honarpardaz has been an industry-based doctoral student working with ABB Corporate Research in Västerås. His research is therefore industry-oriented, and has been part of a larger Horizon 2020 programme, SARAFun, that has just ended. The aim of this programme was to make it possible for even non-experts to construct a robot-based assembly system in one day. This is a task that currently takes at least six months and requires a wide range of expertise.

Mohammadali Honarpardaz, ABB Corporate ResearchMohammadali Honarpardaz. Photo credit: Monica Westman"We nearly managed it. We have cut the time down to two days and a few hours," says Mohammadali Honarpardaz, whose part in the programme was to cut down the time needed to design the robot fingers, the actual gripper.

"The fingers are hugely significant - remember that they are the only contact the robot has with its surroundings. It doesn't matter how well the robot is designed in other aspects if the fingers cannot cope with their task," he says.

Automatic design

In the work presented in the thesis he has developed a method to automatically design robot fingers for different types of task, when dealing with the mounting of small parts. This is a complicated task that requires high precision and repeatability. Instead of using expensive and complex tactile sensors in the finger tips, just a robust and cheap strain gage sensor is integrated into the fingers to measure the load. The robot must carry out the same task repeatedly, which is something that the robot is taught quite simply by physically demonstrating to it what it is to do.

"Designing a simple gripper that opens and closes with a certain force takes three minutes using the framework that I propose in the thesis. Today, this is done manually, and takes at least a month for an expert," he says.

He has also shown that there are no differences in quality between the automatic and manual design of fingers. The small differences that were measured were less than 0.5%. Another conclusion from the thesis is that the method can be used for slightly more complicated robot fingers that can carry out several different tasks with the same design without needing to change fingers.

Industrial needs

More research into robot fingers, however, will be necessary, and the needs of industry are pressing.

"There is a large divide between academic research and research carried out in industry. In the academic sphere, researchers are happy to solve problems they have found themselves, but not the problems that are relevant to industry. The fact that humans have five fingers depends on the need to develop sufficient force in our hands: an individual finger is too weak on its own. Even so, the majority of research in the field deals with anthropomorphic hands, but there are loads of problems left to solve with two-fingered and three-fingered grippers," says Mohammadali Honarpardaz.

And he will be happy to get to grips with this in new projects with researchers at Linköping University.

The thesis:
Finger Design Automation for Industrial Robots, A Generic and Agile Approach, Mohammadali Honarpardaz, Division of Machine Design, Department of Management and Engineering, Linköping University 2018
Dissertation No. 1917. Principal supervisor Professor Johan Ölvander.


News about robotics

Latest news from LiU