Position: Home > News > industry
Mit develops blind robots that can freely pass through dark dangerous areas
Core hints :Robot, do you want it or not? This special series explores the evolving relationship between human beings and machines and studies the impact of robots, artificial intelligence and automation on our work and life.

Release date:2018-07-14

Browse number:78

Robot, do you want it or not? This special series explores the evolving relationship between human beings and machines and studies the impact of robots, artificial intelligence and automation on our work and life.
Boston power's intelligent robots use complex computer vision, but the third generation cheetah robots of Massachusetts institute of technology adopt different strategies.
The third generation of vision-free cheetah robots developed by the Massachusetts institute of technology weigh 80 pounds and are about the size of a Labrador. they can walk through dark rooms without using cameras or environmental sensors, or climb a staircase with obstacles. It relies on what engineers call " blind location" - that is, feedback from its robot legs, and the algorithm-based sense of balance it needs to travel through the dark.
Sangg BAE Kim, associate professor of mechanical engineering at MIT, said at today's press conference: " robots should be able to handle many unexpected behaviors without relying excessively on visual systems. "
Kim said: " vision may have noise, some inaccuracies, and sometimes even total unavailability. If you rely too much on vision, the robot must be positioned very accurately, which will cause it to move very slowly. Therefore, we hope robots rely more on tactile information. In this way, it can handle unexpected obstacles while moving quickly. "
This strategy is very suitable for robots to move in disaster areas or other dangerous environments.
Kim said: " the third-generation cheetah robot was designed to do a variety of tasks. these tasks involve various terrain conditions, including stairs, curbs, and ground covered with obstacles. I think that on many occasions, we will want robots to replace human beings to accomplish simple tasks. Through remote control of robots, dangerous, dirty and difficult work can be done more safely. "
Kim's team developed two new algorithms for cheetah robots without vision systems.
The contact detection algorithm helps the robot determine the best time point to switch between the swing of the leg and stepping on the ground, depending on the obstacle it feels when touching the ground. The algorithm calculates these probabilities based on data from gyroscopes, accelerometers, and leg joint positions.
Kim explained: " if humans close their eyes and take a step forward, we will have a psychological model of where the ground may be and we can be prepared for this. But we also rely on the feeling of touching the ground. What we did for the 3rd generation cheetah robot was similar, combining information from multiple sources to determine the timing of the switch. "
Researchers tested the algorithm in experiments: let the third generation cheetah robot trot on the treadmill in the laboratory and climb stairs. Both surfaces are randomly strewn with some objects, such as wood blocks and tape rolls.
At the same time, a model predictive control algorithm is that when any leg touches the ground and exerts a force of a certain magnitude, the model predictive control algorithm will immediately calculate where the robot's body and legs should be in the next half second.
Kim said: " for example, if someone kicks the robot from the side, when the robot's foot has touched the ground, the algorithm will decide' how should I determine the force to be exerted by the foot? Because there is a speed that hinders me on the left, I have to apply a force to the opposite direction to eliminate that speed. If I apply a force of 100 newtons in the opposite direction, what will happen in half a second? ’”
The prediction algorithm calculates 20 times per second. In order to test its performance, in the experiment, the researchers kicked and pulled the robot with their feet while the robot was on the treadmill or upstairs, and then adjusted the algorithm accordingly. ( we hope that the third generation cheetah robots will not bear grudges against this ).
In the end, Kim and his colleagues will add computer vision systems to the existing ones, but for the third generation cheetah robots, the first thing they want to do is blind eye movement.
A robot that can walk, run, or climb in the dark and is definitely much faster than human beings? This is exactly what you want to see after the earthquake - and this is exactly what you don't want to see after the robot uprising.
At the international conference on intelligent robots and systems to be held in October this year, no vision technology and other improvements to cheetah robot models will become the theme of this conference. The research was supported by naver, Toyota research organization, Foxconn and the us air force scientific research office.

Friendship link    | International Artificial intelligence

News Exhibition Video Product Enterprise Picture Technology
Liaoning ICP No. 234234234 -1 | Liaoning No. 1102342342030919 public security