tasks package

The Task Class

class autompc.tasks.Task(system)

Defines a control task to be solved

set_num_steps(num_steps)

Sets maximum number of steps as the task terminiation condition.

Parameters:

num_steps (int) – Maximum number of steps.

has_num_steps()

Check whether task has a maximum number of steps for the task.

Returns:

True if maximum number of steps is set

Return type:

bool

get_num_steps()

Returns the maxium number steps if available. None otherwise.

term_cond(traj)

Checks the task termination condition.

Parameters:

traj (Trajectory) – Trajectory to check termination condition.

Returns:

True if termination condition met.

Return type:

bool

set_term_cond(term_cond)

Set the task termination condition

Parameters:

term_cond (Function, Trajectory -> bool) – Termination condition function.

set_cost(cost)

Sets the task cost

Parameters:

cost (Cost) – Cost to be set

get_cost()

Get the task cost

Returns:

Task cost

Return type:

Cost

set_init_obs(init_obs)

Sets the initial observation for the task.

Parameters:

init_obs (numpy array of size self.system.obs_dim) – Initial observation

get_init_obs()

Get the initial observation for the task

Returnsnumpy array of size self.system.obs_dim

Initial observation

set_obs_bound(obs_label, lower, upper)

Set a bound for one dimension of the observation

Parameters:

• obs_label (string) – Name of observation dimension to be bounded

• lower (float) – Lower bound

• upper (float) – Upper bound

set_obs_bounds(lowers, uppers)

Set bound for all observation dimensions

Parameters:

• lowers (numpy array of size self.system.obs_dim) – Lower bounds

• uppers (numpy array of size self.system.obs_dim) – Upper bounds

set_ctrl_bound(ctrl_label, lower, upper)

Set a bound for one dimension of the control

Parameters:

• ctrl_label (string) – Name of control dimension to be bounded

• lower (float) – Lower bound

• upper (float) – Upper bound

set_ctrl_bounds(lowers, uppers)

Set bound for all control dimensions

Parameters:

• lowers (numpy array of size self.system.ctrl_dim) – Lower bounds

• uppers (numpy array of size self.system.ctrl_dim) – Upper bounds

are_obs_bounded()

Check whether task has observation bounds

Returns:

True if any observation dimension is bounded

Return type:

bool

are_ctrl_bounded()

Check whether task has control bounds

Returns:

True if any control dimension is bounded

Return type:

bool

get_obs_bounds()

Get observation bounds. If unbounded, lower and upper bound are -np.inf and +np.inf respectively.

Returns:

Observation bounds

Return type:

numpy array of shape (self.system.obs_dim, 2)

get_ctrl_bounds()

Get control bounds. If unbounded, lower and upper bound are -np.inf and +np.inf respectively.

Returns:

Control bounds

Return type:

numpy array of shape (self.system.ctrl_dim, 2)