CARLA integrates the [C++ Library for Responsibility Sensitive Safety](https://github.com/intel/ad-rss-lib) in the client library. This feature allows users to investigate behaviours of RSS without having to implement anything. CARLA will take care of providing the input, and applying the output to the AD systems on the fly.
The RSS library implements a mathematical model for safety assurance. It receives sensor information, and provides restrictions to the controllers of a vehicle. To sum up, the RSS module uses the sensor data to define __situations__. A situation describes the state of the ego vehicle with an element of the environment. For each situation, safety checks are made, and a proper response is calculated. The overall response is the result of all of the combined. For specific information on the library, read the [documentation](https://intel.github.io/ad-rss-lib/), especially the [Background section](https://intel.github.io/ad-rss-lib/ad_rss/Overview/).
The RSS integration has to be built aside from the rest of CARLA. The __ad-rss-lib__ comes with an LGPL-2.1 open-source license that creates conflict. It has to be linked statically into *libCarla*.
There are additional prerequisites required for building RSS and its dependencies. Take a look at the [official documentation](https://intel.github.io/ad-rss-lib/BUILDING) to know more about this.
[__carla.RssSensor__](python_api.md#carla.RssSensor) supports [ad-rss-lib v4.2.0 feature set](https://intel.github.io/ad-rss-lib/RELEASE_NOTES_AND_DISCLAIMERS) completely, including intersections, [stay on road](https://intel.github.io/ad-rss-lib/ad_rss_map_integration/HandleRoadBoundaries/) support and [unstructured constellations (e.g. with pedestrians)](https://intel.github.io/ad-rss-lib/ad_rss/UnstructuredConstellations/).
So far, the server provides the sensor with ground truth data of the surroundings that includes the state of other traffic participants and traffic lights.
When the client calls for it, the [__carla.RssRestrictor__](python_api.md#carla.RssRestrictor) will modify the vehicle controller to best reach the desired accelerations or decelerations by a given response.
Due to the stucture of [carla.VehicleControl](python_api.md#carla.VehicleControl) objects, the restrictions applied have certain limitations. These controllers include `throttle`, `brake` and `streering` values. However, due to car physics and the simple control options these might not be met. The restriction intervenes in lateral direction simply by counter steering towards the parallel lane direction. The brake will be activated if deceleration requested by RSS. This depends on vehicle mass and brake torques provided by the [carla.Vehicle](python_api.md#carla.Vehicle).
!!! Note
In an automated vehicle controller it might be possible to adapt the planned trajectory to the restrictions. A fast control loop (>1KHz) can be used to ensure these are followed.
That sets the basics regarding the RSS sensor in CARLA. Find out more about the specific attributes and parameters in the [sensor reference](ref_sensors.md#rss-sensor).