October 6, 2019
This is the processing of visual information to obtain knowledge. The basic task inside this technology is to detect the object in images and video, i.e. to recognise that one picture in a corner shows a car and the other one shows a computer, keyboard and phone. In robotics, the results of object detection give the robot an understanding of what to do and how to do it, and help it learn.
A logical continuation of detection is tracking, i.e. first the object is detected and then the tracking of its movements begins. Robots need this to understand the visual scene and learn to predict the actions of other objects, which is indispensable, for example, for self-driving cars.
Other tasks for computer vision are segmentation of the image (understanding where the floor is, where the wall is and where the door is) and depth assessment. The latter involves understanding the distance to an object and is solved by reconstructing the three-dimensional geometry from a series of two-dimensional photographs.
How robots help making teleportation real
Natural language processing
Communication with a person is impossible without understanding his or her language. AI specialists disassemble in parts individual morphemes, even the emotional coloration of words in a text, sewing it into a program. Robots need these technologies, for them it’s like a dialogue window with a person, and it’s not just about understanding, but also about responding and learning new concepts.
If the language processing concerns textual information, then the speech analytics is sound. First and foremost, this is speech recognition, which by 2019 has already become solidly mainstreamed into people’s lives. The next step is to speech synthesis and improve the voice quality of the robot and/or program itself to the levels of human communication.
In other words, this technology can be called the automation of processes when they pass without human intervention. Since, again, we are talking about a weak AI that is tailored to individual tasks, the technology for decision-making is perhaps the most understandable in its purpose. The authors of the review highlight several applications for using such technologies:
- navigation, e.g. bypassing obstacles, memorizing and recording the path travelled, localizing yourself in space;
- learning through demonstrations, when the robot memories the actions shown visually or mechanically;
- emotional interaction, for which the machine needs to understand the mood of the person standing in front of you, superimpose it on its “character” features and produce the result as a “mimic” or “emotion”;
- automation of machine learning, i.e. reducing the person’s participation in machine learning, partial transfer to self-learning.
Of course, such technologies must be applied together with others: independent navigation together with computer vision, and emotions together with speech analytics.
This technology is remotely similar to decision making, but analysts have identified it as a separate paragraph. The reason for this is the potential for wide application of recommendation systems in service robotics. We are talking about the supply of goods and services, targeted advertising, a selection of films and music. In the case of robots, the technology can lead to the spread of robotic waiters or sales consultants.