osc position fixes

[snasiriany]

What this does

This PR should fix #562

Running OSC_POSE worked but OSC_POSITION would result in the following error:

Traceback (most recent call last):
  File "/archive/tmp/reproduce_v15.py", line 56, in <module>
    obs, reward, done, info = env.step(action)
  File "/home/soroushn/code/robosuite/robosuite/environments/base.py", line 452, in step
    self._pre_action(action, policy_step)
  File "/home/soroushn/code/robosuite/robosuite/environments/robot_env.py", line 589, in _pre_action
    robot.control(robot_action, policy_step=policy_step)
  File "/home/soroushn/code/robosuite/robosuite/robots/fixed_base_robot.py", line 146, in control
    self.composite_controller.set_goal(action)
  File "/home/soroushn/code/robosuite/robosuite/controllers/composite/composite_controller.py", line 103, in set_goal
    controller.set_goal(action)
  File "/home/soroushn/code/robosuite/robosuite/controllers/parts/arm/osc.py", line 263, in set_goal
    self.goal_ori = Rotation.from_rotvec(action[3:6]).as_matrix()
  File "_rotation.pyx", line 1194, in scipy.spatial.transform._rotation.Rotation.from_rotvec
ValueError: Expected `rot_vec` to have shape (3,) or (N, 3), got (0,)

Issue was that in osc.py there was no logic to handle cases where we don't want to control orientation. Fixed by adding "zero" delta actions for orientation.

How it was tested

Explain/show how you tested your changes.

Ran pytest and ran following script adapted from a user:

import robosuite as suite
import numpy as np
from robosuite.controllers.composite.composite_controller_factory import refactor_composite_controller_config

arm_controller_config = suite.load_part_controller_config(
    default_controller='OSC_POSITION',
    # default_controller='OSC_POSE',
)

options = {}

options['env_name'] = "Stack" #'Door'
options['robots'] = "Panda" #'UR5e'
robot = options['robots'][0] if isinstance(
    options['robots'], list,
) else options['robots']
options['initialization_noise'] = None
options['controller_configs'] = refactor_composite_controller_config(
    arm_controller_config, robot, ['right', 'left'],
)


# arm_controller_config['control_delta'] = False
print(options['controller_configs']["body_parts"].keys())
options['controller_configs']["body_parts"]["right"]['input_type'] = "absolute"
options['controller_configs']["body_parts"]["right"]['input_ref_frame'] = "world"


env = suite.make(
    **options,
    use_camera_obs=False,  # Set to True if you want camera observations
    has_offscreen_renderer=False,
    has_renderer=True,  # Enable on-screen rendering
    control_freq=200,  # Control frequency in Hz
)

obs = env.reset()
env.viewer.set_camera(camera_id=0)


target_move = 0.1  # 0.1 meters along x-axis
target = obs["robot0_eef_pos"] + np.array([target_move, target_move, target_move])
# target = np.concatenate((obs["robot0_eef_pos"], [2.11110234, 2.18042231, 0.16471495]))
# target = [0.46814484, 0.01100895, 0.10936336, 2.11110234, 2.18042231, 0.16471495]

# Run a few steps to reach the target position
for i in range(1000):

    # # Set desired action in absolute frame
    action = np.zeros(4)
    action[:3] = target
    # action = np.zeros(7)
    # action[:6] = target

    # Send action to the environment
    obs, reward, done, info = env.step(action)

    # Render the environment for visualization
    env.render()


# Close the environment
env.close()

SECTION TO REMOVE BEFORE SUBMITTING YOUR PR

Note: Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR. Try to avoid tagging more than 3 people.

Note: Before submitting this PR, please read the contributor guideline.