Ранас Мукминов – Physical AI: От нейросетей к гуманоидным роботам (страница 1)
Ранас Мукминов
Physical AI: От нейросетей к гуманоидным роботам
Глава 1. Концепция Embodied AI
Воплощенный искусственный интеллект (Embodied AI) – это системы, которые способны воспринимать физический мир через сенсоры и взаимодействовать с ним через манипуляторы. В отличие от текстовых моделей, физический ИИ должен понимать гравитацию, инерцию, текстуру материалов и пространственную геометрию. Обучение таких систем требует колоссальных объемов мультимодальных данных: видео с камер, показаний лидаров, гироскопов и датчиков давления на кончиках роботизированных пальцев.
Глава 2. Симуляции и Digital Twins
Обучать робота методом проб и ошибок в реальном мире слишком дорого и опасно. Поэтому обучение (Reinforcement Learning) происходит в виртуальных симуляторах (Digital Twins), созданных на движках типа NVIDIA Omniverse или Unreal Engine. В этих средах агенты проходят миллионы итераций, учась ходить, брать хрупкие предметы или собирать детали. После того как весы модели оптимизированы в симуляции (Sim-to-Real transfer), их загружают в мозг реального гуманоидного робота, который начинает выполнять задачу на реальном заводе.
Приложение: Справочная документация PYTHON3
PYTHON(1) General Commands Manual PYTHON(1)
NAME python – an interpreted, interactive, object-oriented programming language
SYNOPSIS python [ -B ] [ -b ] [ -d ] [ -E ] [ -h ] [ -i ] [ -I ] [ -m module-name ] [ -q ] [ -O ] [ -OO ] [ -s ] [ -S ] [ -u ] [ -v ] [ -V ] [ -W argument ] [ -x ] [ [ -X option ] -? ] [ –check-hash-based-pycs default | always | never ] [ -c command | script | – ] [ arguments ]
DESCRIPTION Python is an interpreted, interactive, object-oriented programming language that combines remarkable power with very clear syntax. For an introduc‐ tion to programming in Python, see the Python Tutorial. The Python Library Reference documents built-in and standard types, constants, functions and modules. Finally, the Python Reference Manual describes the syntax and semantics of the core language in (perhaps too) much detail. (These docu‐ ments may be located via the INTERNET RESOURCES below; they may be installed on your system as well.)
Python's basic power can be extended with your own modules written in C or C++. On most systems such modules may be dynamically loaded. Python is also adaptable as an extension language for existing applications. See the internal documentation for hints.
Documentation for installed Python modules and packages can be viewed by running the pydoc program.
COMMAND LINE OPTIONS -B Don't write .pyc files on import. See also PYTHONDONTWRITEBYTECODE.
–b Issue warnings about str(bytes_instance), str(bytearray_instance) and comparing bytes/bytearray with str. (-bb: issue errors)
–c command Specify the command to execute (see next section). This terminates the option list (following options are passed as arguments to the com‐ mand).
–-check-hash-based-pycs mode Configure how Python evaluates the up-to-dateness of hash-based .pyc files.
–d Turn on parser debugging output (for expert only, depending on compilation options).
–E Ignore environment variables like PYTHONPATH and PYTHONHOME that modify the behavior of the interpreter.
–h , -? , –help Prints the usage for the interpreter executable and exits.
–i When a script is passed as first argument or the -c option is used, enter interactive mode after executing the script or the command. It does not read the $PYTHONSTARTUP file. This can be useful to inspect global variables or a stack trace when a script raises an exception.
–I Run Python in isolated mode. This also implies -E and -s. In isolated mode sys.path contains neither the script's directory nor the user's site-packages directory. All PYTHON* environment variables are ignored, too. Further restrictions may be imposed to prevent the user from in‐ jecting malicious code.
–m module-name Searches sys.path for the named module and runs the corresponding .py file as a script.
–O Remove assert statements and any code conditional on the value of __debug__; augment the filename for compiled (bytecode) files by adding .opt-1 before the .pyc extension.
–OO Do -O and also discard docstrings; change the filename for compiled (bytecode) files by adding .opt-2 before the .pyc extension.
–q Do not print the version and copyright messages. These messages are also suppressed in non-interactive mode.
–s Don't add user site directory to sys.path.
–S Disable the import of the module site and the site-dependent manipulations of sys.path that it entails. Also disable these manipulations if site is explicitly imported later.
–u Force the stdout and stderr streams to be unbuffered. This option has no effect on the stdin stream.
–v Print a message each time a module is initialized, showing the place (filename or built-in module) from which it is loaded. When given twice, print a message for each file that is checked for when searching for a module. Also provides information on module cleanup at exit.
–V , –version Prints the Python version number of the executable and exits. When given twice, print more information about the build.
–W argument Warning control. Python sometimes prints warning message to sys.stderr. A typical warning message has the following form: file:line: cate‐ gory: message. By default, each warning is printed once for each source line where it occurs. This option controls how often warnings are printed. Multiple -W options may be given; when a warning matches more than one option, the action for the last matching option is performed. Invalid -W options are ignored (a warning message is printed about invalid options when the first warning is issued). Warnings can also be controlled from within a Python program using the warnings module.
The simplest form of argument is one of the following action strings (or a unique abbreviation): ignore to ignore all warnings; default to ex‐ plicitly request the default behavior (printing each warning once per source line); all to print a warning each time it occurs (this may gen‐ erate many messages if a warning is triggered repeatedly for the same source line, such as inside a loop); module to print each warning only the first time it occurs in each module; once to print each warning only the first time it occurs in the program; or error to raise an excep‐ tion instead of printing a warning message.
The full form of argument is action:message:category:module:line. Here, action is as explained above but only applies to messages that match the remaining fields. Empty fields match all values; trailing empty fields may be omitted. The message field matches the start of the warn‐ ing message printed; this match is case-insensitive. The category field matches the warning category. This must be a class name; the match test whether the actual warning category of the message is a subclass of the specified warning category. The full class name must be given. The module field matches the (fully-qualified) module name; this match is case-sensitive. The line field matches the line number, where zero matches all line numbers and is thus equivalent to an omitted line number.
–X option Set implementation specific option. The following options are available:
–X faulthandler: enable faulthandler
–X showrefcount: output the total reference count and number of used memory blocks when the program finishes or after each statement in the interactive interpreter. This only works on debug builds
–X tracemalloc: start tracing Python memory allocations using the tracemalloc module. By default, only the most recent frame is stored in a traceback of a trace. Use -X tracemalloc=NFRAME to start tracing with a traceback limit of NFRAME frames
–X importtime: show how long each import takes. It shows module name, cumulative time (including nested imports) and self time (excluding nested imports). Note that its output may be broken in multi-threaded application. Typical usage is python3 -X importtime -c 'import asyncio'
–X dev: enable CPython's "development mode", introducing additional runtime checks which are too expensive to be enabled by default. It will not be more verbose than the default if the code is correct: new warnings are only emitted when an issue is detected. Effect of the developer mode: * Add default warning filter, as -W default * Install debug hooks on memory allocators: see the PyMem_SetupDebugHooks() C function * Enable the faulthandler module to dump the Python traceback on a crash * Enable asyncio debug mode * Set the dev_mode attribute of sys.flags to True * io.IOBase destructor logs close() exceptions
–X utf8: enable UTF-8 mode for operating system interfaces, overriding the default locale-aware mode. -X utf8=0 explicitly disables UTF-8 mode (even when it would otherwise activate automatically). See PYTHONUTF8 for more details