RoSys provides an easy-to-use robot system. Its purpose is similar to ROS. But RoSys is fully based on modern web technologies and focusses on mobile robotics.
The full documentation is available at rosys.io.
Python is great to write business logic. Computation-heavy tasks are wrapped in processes, accessed through WebSockets or called via C++ bindings. Like you would do in any other Python program.
You can structure your code as you please. RoSys provides its magic without assuming a specific file structure, configuration files or enforced naming.
Thanks to asyncio you can write your business logic without locks and mutexes. The execution is parallel but not concurrent which makes it easier to read, write and debug. In real-case scenarios this is also much faster than ROS. Its multiprocessing architecture requires too much inter-process communication.
Most machines need some kind of human interaction. RoSys is built from the ground up to make sure your robot can be operated fully off the grid with any web browser. This is done by incorporating NiceGUI, a wonderful all-Python UI web framework. It is also possible to proxy the user interface through a gateway for remote operation.
Robot hardware is often slower than your own computer. To rapidly test out new behavior and algorithms, RoSys provides a simulation mode. Here, all hardware is mocked and can even be manipulated to test wheel blockages and similar.
You can use pytest to write high-level integration tests. It is based on the above-described simulation mode and accelerates the robot's time for super fast execution.
RoSys modules are just Python modules which encapsulate certain functionality. They can hold their own state, register lifecycle hooks, run methods repeatedly and subscribe to or raise events. Modules can depend on other modules which is mostly implemented by passing them into the constructor.
Modules can register functions via rosys.on_startup or rosys.on_shutdown as well as repeatedly with a given interval with rosys.on_repeat.
!!! note
Note that NiceGUI's app object also provides methods app.on_startup and app.on_shutdown, but it is recommended to use RoSys' counterparts:
rosys.on_startup ensures the callback is executed after persistent modules have been loaded from storage.
If you, e.g., set the rosys.config.simulation_speed programmatically via app.on_startup() instead of rosys.on_startup,
the change is overwritten by RoSys' persistence.restore().
Modules can provide events to allow connecting otherwise separated modules of the system.
For example, one module might read sensor data and raise an event NEW_SENSOR_DATA, without knowing of any consumers.
Another module can register on NEW_SENSOR_DATA and act accordingly when being called.
RoSys provides an Automator module for running "automations".
Automations are coroutines that can not only be started and stopped, but also paused and resumed, e.g. using AutomationControls.
Have a look at our Click-and-drive example.
Modules can register backup and restore methods to read and write their state to disk.
RoSys uses its own time which is accessible through rosys.time.
This way the time can advance much faster in simulation and tests if no CPU-intensive operation is performed.
To delay the execution of a coroutine, you should invoke await rosys.sleep(seconds: float).
This creates a delay until the provided amount of RoSys time has elapsed.
RoSys makes extensive use of async/await to achieve parallelism without threading or multiprocessing. But not every piece of code you want to integrate is offering an asyncio interface. Therefore RoSys provides two handy wrappers:
IO-bound:
If you need to read from an external device or use a non-async HTTP library like requests,
you should wrap the code in a function and await it with await rosys.run.io_bound(...).
CPU-bound:
If you need to do some heavy computation and want to spawn another process,
you should wrap the code in a function and await it with await rosys.run.cpu_bound(...).
Python (and Linux) is fast enough for most high-level logic, but has no realtime guarantees. Safety-relevant behavior should therefore be put on a suitable microcontroller. It governs the hardware of the robot and must be able to perform safety actions like triggering emergency hold etc.
We suggest to use an industrial PC with an integrated controller like the Zauberzeug Robot Brain. It provides a Linux system to run RoSys, offers AI acceleration via NVidia Jetson, two integrated ESP32 microcontrollers and six I/O sockets with up to 24 GPIOs for digital I/Os, CAN, RS485, SPI, I2C, etc. It also has two hardware ENABLE switches and one which is controllable via software.
To have flexible configuration for the microcontroller we created another open source project called Lizard. It is a domain-specific language interpreted by the microcontroller which enables you to write reactive hardware behavior without recompiling and flashing.
RoSys builds upon the open source project NiceGUI and offers many robot-related UI elements. NiceGUI is a high-level UI framework for the web. This means you can write all UI code in Python and the state is automatically reflected in the browser through WebSockets. See any of our examples.
RoSys can also be used with other user interfaces or interaction models if required, for example a completely app-based control through Bluetooth Low Energy with Flutter.
Modules can notify the user through rosys.notify('message to the user').
When using NiceGUI, the notifications will show as snackbar messages.
The history of notifications is stored in the list rosys.notifications.