Autonomous thinking aerial robots flew out of Zhejiang University in droves and appeared on the cover of science sub-issues

Reporting by XinZhiyuan

EDIT: Time

The flight planner independently developed by the mainland is equivalent to installing a smart brain on the UAV, which can not only plan the route independently, but also successfully avoid obstacles.

Recently, in a bamboo forest in Anji, Zhejiang Province, a group of miniature intelligent drones are collectively dispatching, and they are navigating through the jungle with ease.

Look, it's like a military exercise, or a bird walking through the woods.

4.5MB

These robot formations, sometimes skimming the jungle lightly, sometimes crossing and changing formations, like geese flying by, like butterflies dancing.

In fact, this is a study completed by the School of Control Science and Engineering of Zhejiang University and the Zhejiang Dahuzhou Research Institute, which was published in science robotics, a sub-journal of Science in May this year.

Address of the paper: https://www.science.org/doi/10.1126/scirobotics.abm5954

Travel through the jungle

How can drones find the best route on their own?

In the dense forest, for drones, this is the chaotic environment.

In such an environment, it is difficult for a drone to find a route, let alone a group of "group navigation".

Now, the mainland's self-developed planner is equivalent to installing a "smart brain" for the drone, which can not only accurately plan the route, but also be as light as a swallow to cleverly avoid obstacles.

Because autonomous flight in groups in an unknown complex environment has always been regarded as a major technical bottleneck for AI and robot flight.

However, in more than 2 years of research, the Zhejiang University team has worked hard to solve this problem, the core of which is intelligent navigation and fast obstacle avoidance.

Flying in the jungle, flying out, with a healthy body, isn't this the shot in a Chinese martial arts movie?

Alas, people are still too heavy, and it is inevitable to destroy the bamboo forest and destroy the ecology.

Well, what about the miniature drone we developed, it doesn't have a bottle of Coke weight yet!

As a result, these intelligent micro-drone swarms can fly like this:

It can also fly like this:

So, how do these drone swarms achieve intelligent flight?

It's not that all roads lead to Rome, there are obstacles that are not terrible, drones can fly around.

The curve in the middle of the above figure is the flight path planned by the drone after successfully avoiding obstacles.

In fact, there are countless obstacles in the wild, so drones have to complete these obstacle avoidance tasks one by one.

Once these obstacles are identified, the drone can accurately plan its own route.

Even if the future is full of thorns, drones are smart enough to plan their own routes.

Research team

The first author of the paper is Zhou Xin, a doctoral student in the School of Control Science and Engineering of Zhejiang University, and the corresponding authors are Professor Xu Chao and Gao Fei, associate researchers of the institute.

"The herringbone geese, the orderly flocks of pigeons, when we look up at the sky, the flocks of birds in nature always make us think about how they keep their formations." Professor Xu Chao said.

Allowing robots to fly freely in flocks like birds is a natural goal that the research team is struggling to pursue.

"We've built an intelligent brain for the robots in the air." Professor Xu Chao said.

The key to research is to simplify the complexity. Professor Xu Chao believes that "the motion information of the physical world is characterized into a series of mathematical problems, and the hidden special mathematical structures behind complex problems are recognized."

Then, use the on-board computing resources to solve the problem perfectly.

Although the processor is only the size of a thumb, it is capable of independently calculating and processing the massive amounts of information encountered during flight.

A single aerial robot is only the size of a palm and weighs less than a bottle of Coke. Through collective and concerted efforts, the Zhejiang University team developed a small, lightweight, autonomous and controllable flight system that can be used in groups.

"We let robots understand the world, actively avoid obstacles, and achieve group flight." Associate researcher Goofy said.

Using only onboard vision and onboard computing resources, these drones realize the perception of surrounding obstacles in the complex environment of wild woods.

In addition, these UAVs can also locate their own positions and generate flight paths, achieving a number of key technological breakthroughs such as multi-agent communication.

Professor Xu Chao also believes that the research has made breakthroughs in a number of technologies such as intelligence, networking, and autonomy, and this achievement will also have a positive effect on robots and industrial upgrading in the industrial community.

"There are many scenarios in life where we can use this technology, such as sweeping floors or service robots, and if we install the 'smart brain' we developed, it will be smarter." Associate Researcher Goofy said so.

When it comes to future applications, Gao Fei believes that in search and rescue scenarios such as fires, small cluster robots can better achieve search and rescue goals and reduce the risk of search and rescue personnel; in terrain exploration, they can also quickly model areas that cannot be reached by personnel.

Resources:

http://www.news.zju.edu.cn/2022/0505/c24345a2526071/page.htm

http://www.ailab.cn/robot/20220505128808.html