Hand crafted

Software that may give manufacturers a better grip on industrial processes is being developed by researchers at Portsmouth and China’s Shanghai Jiao Tong universities.

It will be used to create a perfect robotic hand capable of recreating the human hand’s range of movement and level of control. The researchers hope it will give a greater level of flexibility and control in manufacturing.

They also hope to integrate a brain computer interface (BCI) to allow the technology to be used as a prosthetic, giving severely disabled people and paraplegics the ability to perform everyday actions.

‘This project is very challenging,’ said Dr Honghai Liu, senior lecturer at Portsmouth University’s Institute of Industrial Research. ‘At the moment we are developing a framework and the data has been very positive. We want to develop a proper algorithm and use it in the hardware.

‘A robotic hand which can perform tasks with the dexterity of a human hand is one of the ‘Holy Grails’ of science. We are talking about having a super high-level control of a robotic device. Nothing which exists today even comes close.’

The team used motion capture to collect data about the way human hands move. A cyber glove, covered in sensors, was filmed by eight high-resolution charge-coupled device cameras with infrared illumination and measurement accuracy of up to a few millimetres.

Using artificial intelligence based on fuzzy qualitative reasoning and spiking neural networks, the software will be able to analyse the movement of the human hand using the data collected. This combination allows a symbolic description of human movement to control modules in a numeric form.

Qualitative reasoning can represent the grasp and manipulation skills of the human hand while the spiking neural network — the so-called third generation of neural networks that increase the level of realism in a neural simulation — can convert the qualitative description for the robotic control modules.

The software will also be able to learn the movements using a parametric knowledge base. Imitation learning and reasoning will be used to train the parameters. The team expects the spiking neural network will be able to store human hand motion habits, which will allow the hardware to learn and copy the complex and subtle movements of a human hand.

With the correct control mechanism Liu believes the technology could be used to help those without full use of their arms regain their independence. ‘Eventually what we would like to see is how we can control it using human brain signals,’ added Liu. ‘We are discussing using a BCI with people at Southampton University. It could then be used to assist people at home for everyday life.’

The technology also has potential in many other fields, including manufacturing and medicine. Delicate operations that require a light touch and fine movements could be automated, enabling robotics to handle fragile objects and complete more complicated tasks.

The hands could be used in a manufacturing environment to perform tasks previously thought too fiddly for machines but without specificity, giving them the versatility to perform a variety of tasks.

‘Humans move efficiently and effectively in a continuous flowing motion, something we have perfected over generations of evolution and which we all learn to do as babies,’ said Prof Xiangyang Zhu from the Robotics Institute at Shanghai Jiao Tong University. ‘Developments in science mean we will all teach robots to move in the same way.’

Although Liu believes the technology has great potential, at this early stage the team is concentrating on the development process with no predicted time scale and no industrial partners yet involved.

‘I cannot give a date,’ said Liu. ‘Though our partner, Prof Zhu, has got good results from this part of the research. We really want to see how good the performance is before getting companies involved. Then we can choose companies based on the importance of the technology.’