The most basic attribute is often the most difficult. Human movements, though they look simple, are quite complex and extremely difficult to replicate. But there have been some improvements recently in robot technology.
A large amount of research is being conducted on developing robots that are capable of mimicking all human movements perfectly. And though it might be a while before this becomes a reality, a group of scientists from OpenAI has unveiled a robotic hand that is capable of moving its hands exactly the way humans do.
The team at OpenAI used a hand created by a London based tech firm called the Shadow Robot Company. The robot learns how to make human-like movements using a technique called Reinforcement Learning (RL) in which the robot is rewarded or punished depending on whether or not it fulfills its assigned task.
In this video, OpenAI researchers train neural networks in a simulation where a robotic hand can solve the Rubik’s Cube:
And instead of following a trial and error method that could end up taking a lot of time, the researchers used a virtual environment to create simulations through which the robot could learn and evolve the right way to mimic human hand movements. “By learning in simulation, we can gather more experience quickly by scaling up, and by de-emphasizing realism, we can tackle problems that simulators can only model approximately,” European Scientist quotes the OpenAI report.
The RL system incentivizes the robot to learn from its mistake. So every time the robot errs in moving a hand in a particular way, it can pinpoint when the error is made and then try out new hand movements that will match with the expectation.
“It’s just doing random things and failing miserably all the time… Then what we do is we give it a reward whenever it does something that slightly moves it toward the goal it actually wants to achieve, which is rotating the block”, Wired quotes Matthias Plappert, an engineer at OpenAI. Until now, the robot hand has mastered around 15 hand movements. This even includes complex actions like spinning a block or reorienting it using gravity.
OpenAI is a non-profit research organization that is funded by Elon Musk and aims to develop safe Artificial General Intelligence (AGI).
Other human movement mimic projects
A research group at Northeastern University is also working toward creating robot hands that can perfectly mimic human movements. The team is focused on enabling the robot hand to master various motions that occur as a result of combining different movements of hands, fingers, wrists, and thumbs.
A unique challenge that the group faces is to teach the robot how to discern when it has gripped an object when to loosen its grip, and so on. “These abilities are already second nature to humans through vision and proprioception (the ability to sense the relative positions of body parts without looking or thinking about it).
Once we’re able to achieve them in robots, they’ll be able to do things like detecting if a grasp is too strong and is squeezing an object too hard”, Taskin Padir Associate Professor at the university writes in an article at The Conversation.
Succeeding in creating human-like robotic hand movements will have a huge impact on various fields. For instance, manufacturing industries could soon start installing robot hands to do dexterous tasks that are currently done by human beings. Similarly, a robot hand could also do complex surgeries and save human lives. And within the next decade or so, it truly is possible that we might see such advanced robots in use everywhere.