Is there vision for a robot without programming?

Question

Python

There are a lot of videos on YouTube, how they teach a typewriter with a camera to circle a brightly colored circle with a square ... There are a lot of buzzwords, code, why this is better than this, etc.
I won’t walk in circles: Maybe there are ready-made solutions that allow, conditionally, a robot on wheels with a Raspberry Pi and a camera connected to it to drive around the apartment and not crash, go around obstacles, see (let there be an icon highlighted from the environment) its charging station and drive in on it, to be able to supplement the functionality with little effort (“robot, that garbage that you drove around by driving into the room is called a ball, roll it to me” well, as an example, it can be simpler). Maybe there is such a thing? Or every manufacturer of a vacuum cleaner robot reinvents the wheel and teaches him to go around a dog and a sofa, well, this is nonsense in our time, to spend resources on such basic, in my opinion, things. I myself, if anything, a helloworld on a python or bash, well, maybe a little more.

Answer 1

Well, this is nonsense in our time, to spend resources on such basic, in my opinion, things. I myself, if anything, a helloworld on a python or bash, well, maybe a little more.

as for me, it’s nonsense to talk like that with the knowledge of “helovord on python or bash”, something +/- is ready-made on github.

"robot, that bullshit that you drove around in the room is called a ball, roll it to me"

Of course, people will add the ability to recognize speech, understand the meaning of words in Russian, and then give it to free access in compiled form and open source.

every robot vacuum manufacturer reinvents the wheel

They can buy software from a company that specializes in it.

Answer 2

Well, this is nonsense in our time, to spend resources on such basic, in my opinion, things. I myself, if anything, a helloworld on a python or bash, well, maybe a little more.

There is a lot of nonsense in this world. And including a huge number of closed corporate developments that are regularly "reinvented" by other corporations. And, moreover, they are presented as a new super-duper-revolutionary. And there is also the risk of "stepping on" someone's patent, like a mine ... A serious discussion of this situation and why it has developed and under what conditions it can be overcome is a serious issue that affects political, social and even philosophical aspects, which I will not go into here.

solutions that allow, conditionally, a robot on wheels with a Raspberry Pi and a camera connected to it to drive around the apartment and not crash, go around obstacles, see (let there be an icon highlighted from the environment) its charging station and drive into it...

Not a long memory and googling gave the following:
1. ROS (Robot operating system):
an open platform that is supposed to be supplemented with modules for what is described in this
paragraph.
- https://www.ros.org/ - off site
- https://en.wikipedia.org/wiki/Robot_Operating_System - wikipedia for quick reference
2. OpenVINO: Some library based on neural networks for common robot vision tasks.
- https://www.intel.com/content/www/us/en/developer/... Off site
- https://habr.com/ru/company/intel/blog/546438/ Article on habré (from students Don't expect much)
Again, for quick reference.
- https://medium.com/the-research-nest/computer-visi... - here, besides OpenVINO,
similar projects
3. integration of the first and second.
- https://github.com/intel/ros_openvino_toolkit

... to be able to supplement the functionality with little effort ("robot, that garbage that you drove around by driving into the room is called a ball, roll it to me" well, as an example, it can be simpler).

This, I'm afraid, is not realistic. Certainly not by Raspberry. Here it may be necessary to reinforce the efforts of the baby with a super-computer quietly noisy in the background. Not from the top 500, of course, but powerful enough, because so far everything is sad with the adaptation of neural networks for new tasks - this is done by a person by restructuring the architecture and learning in a new way (one-short learning is certainly cool, but it solves a simpler task than described ). And how to connect them with the classical algorithms of understanding, logic and planning - I haven’t even seen such works (they probably exist, but marketers don’t shout about them, like about another shit-auto-artist of some savings ...)
PS I don’t a very big specialist in robots, so perhaps these are obvious things or generally worthless projects. But at least that's what open source has.