Deep learning is not a destination

To let the robot can think like a man truly, is not only need to be equipped with kinds of people for its "brain", close to people's cognitive way of the outside world is also indispensable. Alert lady and the flexible are human with good intelligence, robot needs to be central and terminal.

Deep learning is never a destination robot smart still need to strengthen the five senses

Deep learning ability is an important symbol of reflect the degree of intelligent robot, and hence the current major key technology research and development. Through research and simulate human neural network structure and the operation way, and will gradually applied in the design practice scientific research achievements, the depth of the robot to enhance learning ability, intelligence is increasing day by day, and "thinking" is becoming more and more like a man. , however, to let the robot can think like a man truly, is not only need to be equipped with kinds of people for its "brain", close to people's cognition to the outside world is also indispensable.

Human cognition to the outside world comes mainly from the five senses, namely, vision, hearing, smell, taste and touch five senses. Therefore, the current industry in consciousness, give class people to robot is mainly divided into five branches accordingly. Whether the virtual robot or entity robot, the direction of development, as long as it is the kind of person are not open around imitations of the five senses. Only depending on the location and function, and differences in the types of sensory and perception degree.

visual

Humans perceive information about 90% from visual, accordingly, machine vision in the robot also occupies a significant place in the five senses. "Visual organs of robots are photosensitive color sensors, sensor etc.

Last month, the United States at Carnegie Mellon university, college of engineering, the researchers developed a can automatically identify and classify different types of 3 d printing metal powder in the machine vision technology, the accuracy rate of over 95%, is expected to be in five years. Computer powder recognition ability is actually better than well trained people. The researchers think that they would help independent microstructure analysis and research in the future.

auditory

For humans, hearing is second only to the visual senses receive outside information. Right now, hot intelligent voice services, in addition to the robot language learning ability put forward high request, is also a major test of robot hearing. Mainly related to the machine hearing acoustic sensor.

Not content with the iPhone's Siri software can only identify the function of the voice, such as the United States roboticists JosephRomano and his collaborators at the university of Pennsylvania creates a named ROAR (called the robot operating system open source audio identifier) software tools. The software can help roboticists training machine to respond to the voice of the broader significance. Although the technology is still in the initial stage of development, but Romano think its potential is enormous.

The sense of smell

Compared with vision and hearing, smell less widely the application scope of, but for some special purpose robot is crucial for the senses. Gas sensor is to give the sense of smell machine main machine.

In 2015, the university of Tokyo, Japan, sumitomo chemical company and kanagawa institute of science and technology of draw lessons from insect olfactory structure, development feel the sweat of the human body sensors, the device can be installed on the robot, helping them to quickly find the missing, to prevent the secondary disasters for the rescue team. Last year, the team of engineers at the university of Washington in st lewis got inspiration from the locust's sense of smell, developed a new kind of bionic robot sensor system, can be used to sniff out explosives and other dangerous goods.

The sense of taste

Robot need not eat, the industry for the research of machine taste is relatively less. However, in the service of restaurant industry robot, of course, evolved the ability to taste. Thanks to help chemical sensors, robots have the "taste".

Last year, professor of chemistry at the university of valladolid Spain MariaLuzRodriguez - Mendez developed a taster BeerTongue robot. It can have "electronic tongue", through the sensor and chemical methods tasting beer, distinguish rate has reached 82%.

The sense of touch

Touch for entity robot is very important, whether in the factory or in the family, have sense of touch often can improve the quality of service of the robot. A robot tactile mainly through sensors to obtain the pressure sensitive, win min and fluid.

Recently, American scientists at Carnegie Mellon university developed a Fingervision robot system, the interior is equipped with a small camera to act as his fingers to "sense" an object. They are using a desktop 3 d printer to print the holder to act as a robot hands. Fingervision is to let the meaning of huge robot grasping objects, whether can by tactile sensing items appear sliding to control grip strength.

Only with keen sense of five effective gathering outside information, combined with the depth of the strong ability to learn information analysis processing, the degree of intelligent robot can be improved fully. Alert lady and the flexible are human with good intelligence, robot needs to be central and terminal.