smarts.env.gymnasium.driving_smarts_2022_env module

smarts.env.gymnasium.driving_smarts_2022_env.driving_smarts_2022_env(scenario: str, agent_interface: AgentInterface, headless: bool = True, seed: int = 42, visdom: bool = False, sumo_headless: bool = True, envision_record_data_replay_path: str | None = None)[source]

An environment with a mission to be completed by a single or multiple ego agents.

Observation space for each agent:

An unformatted Observation is returned as observation.

Action space for each agent:

Action space for each agent is configured through its AgentInterface. The action space could be either of the following.

  1. RelativeTargetPose

Action

Values

Units

Δx-coordinate

[-2.8, 2.8]

m

Δy-coordinate

[-2.8, 2.8]

m

Δheading

[-π, π]

rad

  1. TargetPose

Action

Values

Units

Next x-coordinate

[-1e10, 1e10]

m

Next y-coordinate

[-1e10, 1e10]

m

Heading with respect to map’s axes

[-π, π]

rad

ΔTime

0.1

s
Reward:

Reward is distance travelled (in meters) in each step, including the termination step.

Episode termination:

Episode is terminated if any of the following occurs.

  1. Steps per episode exceed 800.

  2. Agent collides, drives off road, drives off route, or drives on wrong way.

Parameters:

  • scenario (str) – Scenario name or path to scenario folder.

  • agent_interface (AgentInterface) – Agent interface specification.

  • headless (bool, optional) – If True, disables visualization in Envision. Defaults to False.

  • seed (int, optional) – Random number generator seed. Defaults to 42.

  • visdom (bool, optional) – If True, enables visualization of observed RGB images in visdom. Defaults to False.

  • sumo_headless (bool, optional) – If True, disables visualization in SUMO GUI. Defaults to True.

  • envision_record_data_replay_path (Optional[str], optional) – Envision’s data replay output directory. Defaults to None.

Returns:

An environment described by the input argument scenario.

smarts.env.gymnasium.driving_smarts_2022_env.resolve_agent_action_space(agent_interface: AgentInterface)[source]

Get the competition action space for the given agent interface.

smarts.env.gymnasium.driving_smarts_2022_env.resolve_agent_interface(agent_interface: AgentInterface)[source]

Resolve the agent interface for a given environment. Some interface values can be configured by the user, but others are pre-determined and fixed.

smarts.env.gymnasium.driving_smarts_2022_env.resolve_env_action_space(agent_interfaces: Dict[str, AgentInterface])[source]

Get the environment action space for the given set of agent interfaces.