simulators

BaseSimulator

class core.simulators.base_simulator.BaseSimulator(cfg: Dict)

Base class for simulators.

Arguments:

• cfg (Dict): Config Dict

Interfaces:

apply_control, run_step

abstract apply_control(control: Dict)

Apply control signal to hero vehicle. It will take effect in the next world tick.

Arguments:

control (Dict): Control signal dict

abstract run_step()

Run one step for simulator.

CarlaSimulator

class core.simulators.CarlaSimulator(cfg: Dict, client: Client | None = None, host: str = 'localhost', port: int = 9000, tm_port: int | None = None, timeout: float = 60.0, **kwargs)

Common Carla Simulator. The simulator creates a client to Carla server, and is able to get observation, send control signals to the hero vehicle and record essential data from the simulated world. In the initialization period, the simulator may change the environment parameters including maps and weathers and can add actors (including NPC vehicles, pedestrians as well as sensors mounted on the hero vehicle), During the running period the simulator will achieve running state and information about the hero vehicle (such as running speed, angle, navigation goal and reference path), data from the sensors (such as camera images, lidar points) as well as running status(including collision, running off road, red light, distance and timeout to end waypoint).

Once it is created, it will set up Carla client and set the parameters in the configuration dict as its default. When actually calling the init method to start an episode, some of the configurations may be changed by the input arguments while others remain by default.

The simulator stores and gets some information from a static class CarlaDataProvider to avoid frequently sending message to Carla server and speed up.

Up to now, it uses Carla version higher than 0.9.9.

If no traffic manager port is provided, it will find random free port in system.

Arguments:

• cfg (Dict): Config Dict.

• client (carla.Client, optional): Already established Carla client. Defaults to None.

• host (str, optional): TCP host Carla client link to. Defaults to ‘localhost’.

• port (int, optional): TCP port Carla client link to. Defaults to 9000.

• tm_port (int, optional): Traffic manager port Carla client link to. Defaults to None.

• timeout (float, optional): Carla client link timeout. Defaults to 60.0.

Interfaces:

init, get_state, get_sensor_data, get_navigation, get_information, apply_planner, apply_control, run_step, clean_up

Properties:

• town_name (str): Current town name.

• hero_player (carla.Actor): hero actor in simulation.

• collided (bool): Whether collided in current episode.

• ran_light (bool): Whether ran light in current frame.

• off_road (bool): Whether ran off road in current frame.

• wrong_direction (bool): Whether ran in wrong direction in current frame.

• end_distance (float): Distance to target in current frame.

• end_timeout (float): Timeout for entire route provided by planner.

• total_distance (float): Distance for entire route provided by planner.

apply_control(control: Dict | None = None) → None

Apply control signal for hero player in simulator. This will send message to the client and the control takes effect in next tick

Arguments:

• control (dict, optional): Control signal dict. Default to None.

apply_planner(end_idx: int) → Dict

Apply goal waypoint to planner in simulator. The start point is set to current hero vehicle waypoint.

Arguments:

• end_idx (int): Index of end waypoint.

Returns:

Dict: [description]

clean_up() → None

Destroy all actors and sensors in current world. Clear all messages saved in simulator and data provider. This will NOT destroy the Carla client, so simulator can use same carla client to start next episode.

get_information() → Dict

Get running information including time and ran light counts in current world.

Returns:

Dict: Information dict.

get_navigation() → Dict

Get navigation info in current world. Most of the contains come from planner.

Returns:

Dict: Navigation dict.

get_sensor_data() → Dict

Get all sensor data and bird-eye view data if exist in current world. Bird-eye view will be converted to an multi-channel image.

Returns:

Dict: Sensor and Bev data dict.

get_state() → Dict

Get running state from current world. It contains location, orientation, speed, acc, and the state of surrounding road info such as traffic light and junction.

Returns:

Dict: State dict.

init(start: int = 0, end: int = 1, **kwargs) → None

Init simulator episode with provided args. This method takes start and end waypoint indexes to set a navigation goal, and will use planner to build a route to generate target waypoint and road options in each tick. It will set world, map, vehicles, pedestrians dut to default config, and provided args, which will be stored to replace old config. If no collision happens when creating actors, the init will end and return.

Arguments:

• start (int, optional): Index of start waypoint. Defaults to 0.

• end (int, optional): Index of end waypoint. Defaults to 1.

Optional arguments:

town, weather, n_vehicles, n_pedestrians, autopilot, disable_two_wheels

run_step() → None

Run one step simulation. This will tick Carla world and update information for all sensors and measurement.

Warning

Make sure to use clean_up before delete an simulator. If not, all current actors(vehicles, pedestrians, sensors ect.) will remain in Carla client, and may cause error when next simulator link to that client!

CarlaScenarioSimulator

class core.simulators.CarlaScenarioSimulator(cfg: Dict, client: Client | None = None, host: str = 'localhost', port: int = 9000, tm_port: int = 9050, timeout: float = 60.0, **kwargs)

Simulator used to run scenarios in Carla. The simulator loads the config instance of the provided scenario to start simulation. It can create hero actor, NPC vehicles, walkers, world map according to it and the configuration dict. The modification of sensors and planners, and the running status and information you can get from it are set the same as CarlaSimulator.

When created, it will set up Carla client and store default configuration the same as CarlaSimulator, and it can also be change by the input arguments of the init method.

If no traffic manager port is provided, it will find random free port in system.

Arguments:

• cfg (Dict): Config Dict.

• client (carla.Client, optional): Already established Carla client. Defaults to None.

• host (str, optional): TCP host Carla client link to. Defaults to ‘localhost’.

• port (int, optional): TCP port Carla client link to. Defaults to 9000.

• tm_port (int, optional): Traffic manager port Carla client link to. Defaults to None.

• timeout (float, optional): Carla client link timeout. Defaults to 60.0.

Interfaces:

init, get_state, get_sensor_data, get_navigation, get_information, apply_control, run_step, clean_up

Properties:

• town_name (str): Current town name.

• hero_player (carla.Actor): hero actor in simulation.

• collided (bool): Whether collided in current episode.

• end_distance (float): Distance to target in current frame.

• end_timeout (float): Timeout for entire route provided by planner.

• total_distance (float): Distance for entire route provided by planner.

• scenario_manager (Any): Scenario Manager instance used to get running state.

clean_up() → None

Destroy all actors and sensors in current world. Clear all messages saved in simulator and data provider, and clean up running scenarios. This will NOT destroy theCarla client, so simulator can use same carla client to start next episode.

end_scenario() → None

End current scenario. Must be called before ending an episode.

get_criteria() → List

Get criteria status list of scenario in current frame. Criteria related with hero actor is encounted.

Returns:

List: Criteria list of scenario.

init(config: Any) → None

Init simulator episode with provided args. This method takes an scenario configuration instance to set up scenarios in Carla server. the scenario could be a single scenario, or a route scenario together with several scenarios during navigating the route. A scenario manager is used to manager and check the running status and tick scenarios. A local planner is set to trace the route to generate target waypoint and road options in each tick. It will set world, map, vehicles, pedestrians dut to provided args and default configs, and reset running status. If no collision happens when creating actors, the init will end and return.

Arguments:

• config (Any): Scenario configuration instance, containing information about the scenarios.

run_step() → None

Run one step simulation. This will tick Carla world and scenarios, update information for all sensors and measurement.