There is a pile of puzzles hidden behind the automation revolution.
The human society is experiencing a far-reaching technological re-engineering. Machines that used to play only in science fiction have begun to infiltrate our lives. Even if you don’t have a robotic co-worker, that future is not far off. Autopilot cars are aiming to transform our roads. The first truly advanced robots have been living in hospitals, construction sites and even Wal-Mart.
But behind the automation revolution, there are hidden problems. Being optimistic can also be said to be a challenge. In response, a group of robotics published a paper in the journal Science, Robotics, which listed the top ten challenges facing robot development, covering many areas: new engines from electrical engineering, new materials from materials science, And even the ethical principles of social science. The exact direction of the robot revolution is not clear, but what is certain is that a large number of disciplines will be affected.
“We want to use this article as an entry point for such a diverse field of research, so that everyone can work together to expand their thinking,” said Yang Guangzhong, the first author of the article and imperialist at Imperial College.
We might as well start from the hardware.
The expert group is not concerned with specific robot type issues, such as humanoid robots or collaborative robots. “This is intentional,” said Yang Guangzhong. “Because sometimes we pay too much attention to specific forms, instead of fundamentally thinking about how to find another way, how to learn from nature, how to use new materials.”
Most of the robots are still stupid and expressionless, and it is difficult to speak. But this situation has begun to change. For example, some robot images are getting more and more cute. In the field of the name “soft robot”, engineers are developing a soft robot, such as changing the shape by flowing oil. Such a robot is much safer when used. But first, engineers must overcome some difficulties, such as how the soft robot should repair itself after being punctured. At present, only one kind of soft robot can do it, but only if it is hot for 40 minutes. Ideally, at room temperature, the robot can repair itself.
will be inspired by the nature of the test, this process is called biomimicry. In the field of biological mimicry, self-repair is especially important. For example, to copy a person’s hand, you need a soft material that feels good, but you can repair it yourself after it is damaged. It’s not over yet: the hand is a complex tool full of muscles, tendons and small bones. How to copy it and make the robot hand as smart as a human hand? It seems that the gourd does not work. Therefore, the difficulty lies in how to avoid this complexity while achieving the dexterity comparable to the human hand.
Another example of a biological mimic is the robot “Cassie.” It looks like two legs, like a bird without a body. The most interesting thing is that the creators of Cassie didn’t say, “Okay, copy the pair according to the shape of the bird’s legs.” Instead, by calculation, the most efficient solution is obtained, and the result is exactly similar. Bird legs. But the challenge of copying nature, especially for humanoid robots, is to replicate the unparalleled energy efficiency of living organisms. If you want to simulate an ant-sized robot and drive it with a traditional motor—that is, a super heavy actuator—I can only wish you good luck.
One solution is to think of the robot as part of a distributed whole, not a self-contained executor. For example, micro-robots work together to build complex buildings, or agricultural robots collaborate to harvest crops. In this way, we have to find a way to design a robot that is only a few centimeters long.
Of course, to build interconnected micro-robots, you need to develop algorithms to coordinate hundreds of machines, which involves robotics in the software field. The challenge. In the purely digital world, AI is making great strides, but implementation is another matter.
For example, an algorithm can quickly master certain skills through self-study. For example, after continuous trial and error, learn to recognize objects. This process is called reinforcement learning. But let the robot do it yourself and try the toddler game? Compared to pure virtual computing, this trial and error process is much longer. Therefore, the problem in the future is to let the robot manipulate the novelty objects in the real world.
As the performance of sensors such as Lidar continues to increase and prices are decreasing, robots have made great strides in these areas. But they often fall into the fountain, or almost grind to a dog on the sidewalk. Therefore, this aspect needs to be improved.
This has not yet mentioned the human machine Let’s go together. The most interesting challenge facing robots may involve human-computer interaction. It seems simple – to avoid each other, the robot has to fall to help, and so on – but it will become very tricky at once. For example, last year, a security robot was in trouble, and it was accused of harassing the tramp in San Francisco.
So what is the relationship between us and the machine in emotional and actual interaction? What should we do if producers use these relationships to persuade children to keep buying more and more advanced toy robots? If the robot does not have the ability to respond to love, would you really treat it? From an ethical perspective, the future of robots is inevitably confusing.
Of course, most of the human-computer interaction is harmless and will become the norm in many industries. Some robots, such as the Da Vinci system, have begun to work side by side with surgeons. But the challenge in the future is that as robots increasingly take over boring tasks such as wounds, more and more responsibilities will be handed over to the robot. By that time, human-computer interaction really became a subtle dance, and the surgeons worked side by side with the robots without disturbing each other.
I really want to say that in the near future, who can learn from robotics?
“Look at these top ten challenges,” said Yang Guangzhong. “There are materials from materials science, electricity from electrical engineering, navigation control from computational science, hardware systems from biology.” Ethicists, neuroscientists, brain-computer interfaces, and security experts—prevent your anthropoid robots from getting hacked… It’s more and more like everyone’s chores.
Editor: Man Qian
: Theatre-style offline speech platform, discovering the most creative ideas
Click on the blue word “Learn More” to get more “making” exciting content.