Core

Dynamics models for vehicle simulation.

Provides vehicle dynamics models organized by type. Each model has its own submodule with parameters, dynamics functions, loss functions, and metadata describing how it interacts with the optimisation stack.

Examples:

>>> from pct.dynamics import MODEL_SPECS
>>> spec = MODEL_SPECS["singletrack-augmented"]
>>> params = spec.build_base_params(timestep_hundredths=5)
>>> dynamics_fn = spec.dynamics_fn

ModelSpec dataclass

Metadata describing a dynamics model and its learnable parameters.

Attributes:

Name	Type	Description
name	str	Identifier for the model.
param_keys	tuple of str	Ordered parameter names.
param_shapes	Mapping[str, Tuple[int, ...]]	Shapes for each parameter.
param_bounds	Sequence[Sequence[object]]	Bounds for each parameter.
default_params	Dict[str, object]	Default parameter values.
state_indices	Tuple[int, ...]	State indices used by the model.
dynamics_module	object	Module containing the dynamics implementation.
dynamics_fn	Callable[[object, object], object]	Dynamics function handle.
loss_fn	Callable[..., object]	Loss function handle.
build_base_params	Callable[[int], object]	Factory for baseline parameter objects.
control_dim	int	Control dimensionality.
state_names	Tuple[str, ...]	Human-readable state names.
control_names	Tuple[str, ...]	Human-readable control names.
train_loss_kwargs	Mapping[str, object]	Extra kwargs for training loss.
eval_loss_kwargs	Mapping[str, object]	Extra kwargs for evaluation loss.

full_shapes property

full_shapes

Parameter shapes keyed by parameter name.

vector_size property

vector_size

Flattened parameter vector length.

register_model_spec

register_model_spec(spec)

Register a model specification for global discovery.