📜 ⬆️ ⬇️

How we did the homework robot

Humanity has long dreamed of home robots. Over the past hundred years, the image of a home robot has been formed in science fiction films and books: most often it is an anthropomorphic device with a head, arms and legs. Such a robot is always ready to prepare and bring a cup of coffee to the owner, sweep the floor, wipe the window and do other homework. And still house fiction robots can always talk to the owner. Do not just execute commands, but be a secretary, assistant, partner or even a friend!

image

When my friends and I decided to make our home robot, we also looked at the androids first. I wanted to build a sort of butler so that I could cook breakfast, vacuum the carpet and meet the guests. Unfortunately, after studying the market, we realized that today such a robot can not be done. Rather, you can do something, but it will cost like a new foreign car and require regular service, while adapting such a robot for each specific kitchen will be no easier than rebuilding Ubuntu on the Z80. It is clear that there will not be very many people willing to buy such a robot.
')
And then the understanding came that the housework robot at first does not need to go around the house at all. The inspiration for us was Jarvis (Jarvis) - the hero of the film "Iron Man". This is a butler, an artificial intelligence that does not have a body, but is able to control various devices in the house of the main character. As a result, we have a Cubic - a home robot that lives at home and can fully (more or less) talk with its inhabitants, as well as control various “smart home” systems.



Cubic is a cube that can listen and speak. We printed the early prototype on a 3D printer. It must be placed somewhere in the house or office, plug into the outlet and it is ready to work.
image

Our robot is constantly listening to the surrounding space and waiting for someone to turn to it. Hearing the activator phrase, which is an appeal, he will go into speech recognition mode and listen to what you say to him. Then the cube will process what was said, select the appropriate answer and voice it using a speech synthesizer. It looks like Siri, but in a box :)

The main differences from the Siri - is the work of 24/7, voice recognition at a long distance and fully voice mode. The cube does not have a screen, so it gives all the answers only by voice, and also has a tendency to conduct dialogues and clarifying questions in places where mobile assistants display content on the screen.

Now the cube has about 150 functions - it can read news, quote Wikipedia, as well as many other dictionaries and reference books from astronomical to culinary. The cube can command a smart home (at the embryo level so far), set reminders and alarm clocks, warn about weather phenomena and traffic jams. There is even a voice RPG! For life with a cube, you can also chat, but he knows so many geek tales that he regularly disagrees on quotes. It was not easy to manage with so many functions, so we came up with a whole concept of a voice operating system, which allows you to systematize all the programs, but at the same time speak with the device at ease, and not as IVR.

Of course, this is not a real AI - the cube does not know much, and it is not at all difficult for a person to get him into a dead end. All this is multiplied by speech recognition systems that work far from perfect. But nevertheless it is interesting to talk with him! It often happens that we, the creators, cannot understand whether the AI ​​responds to us or someone connects and “trolls” us in the manual control mode (such an opportunity was created for debugging), and then it is nice to know that the cube responded myself.

I want to show some videos with examples of the cube work:





The development of the device from our friendly company took a year (the three of us started, now there are seven of us). That, by the way, was one of the earliest prototypes.

A year ago, we still did not fully understand what we wanted to do. We thought about the option of a child's robot toys. Subsequently, this idea was abandoned, but the video is still funny. (If someone does not know, this is Pin - the hero of the cartoon about Smeshariki. He should speak with a German accent).



In the end, stopped at the cube.

From a technical point of view, our prototypes are not too complicated. The basis of the device is a minicomputer on Android (it allows us to use Google's speech recognition freely), a speaker, microphones, a radio transmitter for controlling sockets and a backlight. External components are controlled by Arduino. Our main development is a complex system of microphones that allows the cube to be heard at a great distance, up to 10 meters ideally. And, of course, the "intellect" itself.

It is planned that the cube will eventually have an API through which the owner can use our intelligence as the basis for their development. For example, I tried to make a computer of the cabin of a spaceship with a torpedo installation.


In total, we made several prototypes - 10 pieces (despite the simplicity of the device, it turned out to be very difficult!). There are already buyers for these devices. Now we are engaged in the manufacture of an industrial batch of 100 devices (industrial design will be much more beautiful, and DIY components will disappear from the system).

If the topic of our robot turns out to be interesting, then I will write a series of articles about him. How the microphone system works, what intelligence can do, how we put our intellect in the car and started it on smart watches and in general what it is like to make a new gadget in Russia.

Source: https://habr.com/ru/post/215555/


All Articles