Configuration Classes

Dataclass configurations for environments and components.

Module version: 0.2.0

EnvironmentConfig

@dataclass
class EnvironmentConfig:
    """
    Complete configuration for a coopetition environment.

    Used internally by environment classes. Most users should use
    factory functions (make, make_parallel, make_aec) instead.
    """

    n_agents: int = 2
    max_steps: int = 100
    endowments: Optional[NDArray[np.floating]] = None
    alpha: Optional[NDArray[np.floating]] = None
    interdependence_matrix: Optional[NDArray[np.floating]] = None
    value_params: Optional[ValueFunctionParameters] = None
    trust_params: Optional[TrustParameters] = None
    trust_enabled: bool = True
    baselines: Optional[NDArray[np.floating]] = None
    reward_type: str = "integrated"
    normalize_rewards: bool = False
    reward_scale: float = 1.0
    render_mode: Optional[str] = None

Attributes:

Name	Type	Default	Description
`n_agents`	`int`	2	Number of agents
`max_steps`	`int`	100	Maximum episode length
`endowments`	`NDArray`	None	Agent endowments (auto-initialized if None)
`alpha`	`NDArray`	None	Bargaining shares (auto-initialized if None)
`interdependence_matrix`	`NDArray`	None	D matrix (auto-initialized if None)
`value_params`	`ValueFunctionParameters`	None	Value function config
`trust_params`	`TrustParameters`	None	Trust dynamics config
`trust_enabled`	`bool`	True	Enable trust dynamics
`baselines`	`NDArray`	None	Cooperation baselines
`reward_type`	`str`	“integrated”	‘private’, ‘integrated’, or ‘cooperative’
`normalize_rewards`	`bool`	False	Normalize rewards to [0, 1]
`reward_scale`	`float`	1.0	Reward scaling factor
`render_mode`	`str`	None	Rendering mode