generators module

class dynamapp.generators.ModelDataGenerator(model: Model, trajectory: Trajectory)

Bases: object

This class generates data based on a multibody system for a given trajectory. It computes the joint positions, velocities, accelerations, and torques based on the defined trajectory and model dynamics.

Args:

• model (Model): The multibody model to be used.

• trajectory (Trajectory): The trajectory to generate data for.

generate_trajectory_data()

Generate full data for the trajectory with respect to the model's kinematics and dynamics.

Returns:

• dict: A dictionary containing joint positions, velocities, accelerations, and torques

  over the trajectory time.

compute_velocities(q_data)

Compute the joint velocities (q_dot) using the trajectory data.

Args:

• q_data (jnp.ndarray): Joint positions over the trajectory time.

Returns:

• jnp.ndarray: Joint velocities (q_dot).

compute_accelerations(q_data, q_dot_data)

Compute the joint accelerations (q_ddot) using the joint position and velocity data.

Args:

• q_data (jnp.ndarray): Joint positions over the trajectory time.

• q_dot_data (jnp.ndarray): Joint velocities over the trajectory time.

Returns:

• jnp.ndarray: Joint accelerations (npoints * ndof).

compute_torques(q_data, q_dot_data, q_ddot_data)

Compute the joint torques using the Recursive Newton-Euler Algorithm (RNEA).

Args:

• q_data (jnp.ndarray): Joint positions over the trajectory time.

• q_dot_data (jnp.ndarray): Joint velocities over the trajectory time.

• q_ddot_data (jnp.ndarray): Joint accelerations over the trajectory time.

Returns:

• jnp.ndarray: Joint torques (tau).

class dynamapp.generators.ModelStateDataGenerator(model_state: ModelState, num_samples: int, time_steps: int, noise_magnitude: float = 0.1, u_init: jax.numpy.ndarray = None, x_init: jax.numpy.ndarray = None)

Bases: object

Class to generate data for state-dependent multijoint system models.

This class simulates the ModelState class over time and generates a sequence of state, input, and output data.

Args:

• model_state - instance of the ModelState class

• num_samples - number of samples to generate

• time_steps - number of time steps per sample

• noise_magnitude - magnitude of noise to add to the output

• u_init - initial input vector

• x_init - initial state vector

Examples:

>>> model_state = ModelState(Imats, dhparams)
>>> data_generator = ModelStateDataGenerator(model_state, num_samples=100, time_steps=200)
>>> x_data, u_data, y_data = data_generator.generate_data()

Attributes:

• model_state - instance of the ModelState class

• num_samples - number of samples to generate

• time_steps - number of time steps per sample

• noise_magnitude - magnitude of noise to add to the output

• u_init - initial input vector

• x_init - initial state vector

generate_data() → Tuple[jax.numpy.ndarray, jax.numpy.ndarray, jax.numpy.ndarray]

Generate data for a multijoint system, consisting of input (u), state (x), and output (y).

Args:

None

Returns:

• x_data : jnp.ndarray (num_samples, time_steps, 2*ndof) state data

• u_data : jnp.ndarray (num_samples, time_steps, ndof) input data

• y_data : jnp.ndarray (num_samples, time_steps, ndof) output data

compute_data_statistics(x_data: jax.numpy.ndarray, u_data: jax.numpy.ndarray, y_data: jax.numpy.ndarray) → dict

Compute some statistics (mean, std) over the generated data.

Args:

x_data : jnp.ndarray (num_samples, time_steps, 2*ndof) state data u_data : jnp.ndarray (num_samples, time_steps, ndof) input data y_data : jnp.ndarray (num_samples, time_steps, ndof) output data

Returns:

• stats : dict - statistics (mean, std) for state, input, and output data