Blogger Themes

Wednesday 4 April 2012

Robots more human-like than ever are all set to invade Edinburgh

Engineerblogger
April 4, 2012


Professor Sethu Vijayakumar

Robots that think smarter, jump higher and run faster than ever before are one of the big attractions at this year's Edinburgh International Science Festival.

Among the exhibits on show between 04 and 07 April at the Festival's InMotion event will be dancing mini-humanoids and robots that can play Connect Four with the public. There will also be a softball-pitching robot arm and the world's first prototype two-legged walking robot with controllable stiffness, which moves eerily like a human.

The world-leading work on robotics highlighted through these exhibits is being carried out at the University of Edinburgh with funding from the Engineering and Physical Sciences Research Council (EPSRC). The research is helping to usher in an era where robots not only move more like humans but are also able, for the first time, to interpret the context of their actions and self-improve by continuously learning from their experiences just like humans do.

The walking robot incorporates a number of world firsts: the hips, knees and ankles each have three degrees of freedom, meaning that stiffness, 'damping' (the amount of resistance experienced) and position can all be varied simultaneously. In terms of enabling a robot to walk energy efficiently like a human, this is important because, for example, variable damping and stiffness will equip it to deal with differences in terrain and unexpected disturbances, as well as to undertake energetic tasks like jumping, throwing, hopping and running.

Traditionally, robots have been programmed to repeat the same precise movements again and again, working in extremely predictable and repeatable environments. However, the versatile robots on display in Edinburgh will demonstrate their ability to take decisions autonomously and deal with complex situations and unexpected events in a way that robots have never previously been capable of. Such capabilities would be imperative for successful deployment of robots in unpredictable disaster recovery scenarios (e.g. earthquakes, fires, mine collapses) or where communication delays are unavoidable (e.g. interplanetary or deep-sea operations).

"Our work is focused on developing robots that offer the best of both worlds - the reliability of automation combined with the flexibility and sophistication of human decision-making," says Sethu Vijayakumar, Professor of Robotics at the University of Edinburgh, who has led the development of these more human-like robots. "We're successfully addressing the huge computational challenges involved in progressing robotics to the next level by leveraging the success of data-driven machine learning techniques. The result is more precise, more powerful and safer robots that draw inspiration from the way humans think and move."

One area where the work of Professor Vijayakumar's team is showing considerable promise is in the development of prosthetics for humans. 'Wearable robots', functioning as hands or lower legs, for instance, could be controlled through the wearer's muscle activation while also being able to respond autonomously to the environment they are in (e.g. gripping objects that are heavier than expected, or putting down a cup of coffee if it is too hot). Visitors to the Festival will be able to attend a 45-minute, hands-on workshop called 'Make a Move' on the programming of robotic arms.

"The aim is to develop robotic technologies that can cope like humans with 'open-ended environments' - unstructured situations that necessitate the constant ability to learn, adapt and change," Professor Vijayakumar comments. "Within a few years, we could start to see the deployment of robotic devices that interact with and learn autonomously from the environments around them and, perhaps, mimic human problem-solving capabilities while maintaining the benefits of automation such as accuracy and repeatability."


Source: EPSRC


Additional Information:

0 comments:

Post a Comment