mirror of
https://github.com/huggingface/lerobot.git
synced 2026-06-04 04:41:24 +00:00
af9ddcf9a2
* docs(processor): update docstrings batch_processor * docs(processor): update docstrings device_processor * docs(processor): update docstrings tokenizer_processor * update docstrings processor_act * update docstrings for pipeline_features * update docstrings for utils * update docstring for processor_diffusion * update docstrings factory * add docstrings to pi0 processor * add docstring to pi0fast processor * add docstring classifier processor * add docstring to sac processor * add docstring smolvla processor * add docstring to tdmpc processor * add docstring to vqbet processor * add docstrings to converters * add docstrings for delta_action_processor * add docstring to gym action processor * update hil processor * add docstring to joint obs processor * add docstring to migrate_normalize_processor * update docstrings normalize processor * update docstring normalize processor * update docstrings observation processor * update docstrings rename_processor * add docstrings robot_kinematic_processor * cleanup rl comments * add docstring to train.py * add docstring to teleoperate.py * add docstrings to phone_processor.py * add docstrings to teleop_phone.py * add docstrings to control_utils.py * add docstrings to visualization_utils.py --------- Co-authored-by: Pepijn <pepijn@huggingface.co>
105 lines
4.6 KiB
Python
105 lines
4.6 KiB
Python
# !/usr/bin/env python
|
|
|
|
# Copyright 2025 The HuggingFace Inc. team. All rights reserved.
|
|
#
|
|
# Licensed under the Apache License, Version 2.0 (the "License");
|
|
# you may not use this file except in compliance with the License.
|
|
# You may obtain a copy of the License at
|
|
#
|
|
# http://www.apache.org/licenses/LICENSE-2.0
|
|
#
|
|
# Unless required by applicable law or agreed to in writing, software
|
|
# distributed under the License is distributed on an "AS IS" BASIS,
|
|
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
# See the License for the specific language governing permissions and
|
|
# limitations under the License.
|
|
|
|
from dataclasses import dataclass, field
|
|
|
|
from lerobot.configs.types import FeatureType, PolicyFeature
|
|
from lerobot.constants import ACTION
|
|
from lerobot.processor import ActionProcessorStep, ProcessorStepRegistry
|
|
from lerobot.teleoperators.phone.config_phone import PhoneOS
|
|
|
|
|
|
@ProcessorStepRegistry.register("map_phone_action_to_robot_action")
|
|
@dataclass
|
|
class MapPhoneActionToRobotAction(ActionProcessorStep):
|
|
"""
|
|
Maps calibrated phone pose actions to standardized robot action inputs.
|
|
|
|
This processor step acts as a bridge between the phone teleoperator's output
|
|
and the robot's expected action format. It remaps the phone's 6-DoF pose
|
|
(position and rotation) to the robot's target end-effector pose, applying
|
|
necessary axis inversions and swaps. It also interprets platform-specific
|
|
button presses to generate a gripper command.
|
|
|
|
Attributes:
|
|
platform: The operating system of the phone (iOS or Android), used
|
|
to determine the correct button mappings for the gripper.
|
|
"""
|
|
|
|
platform: PhoneOS
|
|
_enabled_prev: bool = field(default=False, init=False, repr=False)
|
|
|
|
def action(self, act: dict) -> dict:
|
|
"""
|
|
Processes the phone action dictionary to create a robot action dictionary.
|
|
|
|
Args:
|
|
act: The input action dictionary from the phone teleoperator.
|
|
|
|
Returns:
|
|
A new action dictionary formatted for the robot controller.
|
|
|
|
Raises:
|
|
ValueError: If 'pos' or 'rot' keys are missing from the input action.
|
|
"""
|
|
# Pop them from the action
|
|
enabled = bool(act.pop(f"{ACTION}.phone.enabled", 0))
|
|
pos = act.pop(f"{ACTION}.phone.pos", None)
|
|
rot = act.pop(f"{ACTION}.phone.rot", None)
|
|
inputs = act.pop(f"{ACTION}.phone.raw_inputs", {})
|
|
|
|
if pos is None or rot is None:
|
|
raise ValueError("pos and rot must be present in action")
|
|
|
|
rotvec = rot.as_rotvec() # Absolute orientation as rotvec
|
|
|
|
# Map certain inputs to certain actions
|
|
if self.platform == PhoneOS.IOS:
|
|
gripper = float(inputs.get("a3", 0.0))
|
|
else:
|
|
a = float(inputs.get("reservedButtonA", 0.0))
|
|
b = float(inputs.get("reservedButtonB", 0.0))
|
|
gripper = (
|
|
a - b
|
|
) # Positive if a is pressed, negative if b is pressed, 0 if both or neither are pressed
|
|
|
|
# For some actions we need to invert the axis
|
|
act[f"{ACTION}.enabled"] = enabled
|
|
act[f"{ACTION}.target_x"] = -pos[1] if enabled else 0.0
|
|
act[f"{ACTION}.target_y"] = pos[0] if enabled else 0.0
|
|
act[f"{ACTION}.target_z"] = pos[2] if enabled else 0.0
|
|
act[f"{ACTION}.target_wx"] = rotvec[1] if enabled else 0.0
|
|
act[f"{ACTION}.target_wy"] = rotvec[0] if enabled else 0.0
|
|
act[f"{ACTION}.target_wz"] = -rotvec[2] if enabled else 0.0
|
|
act[f"{ACTION}.gripper"] = gripper # Still send gripper action when disabled
|
|
return act
|
|
|
|
def transform_features(self, features: dict[str, PolicyFeature]) -> dict[str, PolicyFeature]:
|
|
features.pop(f"{ACTION}.phone.enabled", None)
|
|
features.pop(f"{ACTION}.phone.pos", None)
|
|
features.pop(f"{ACTION}.phone.rot", None)
|
|
features.pop(f"{ACTION}.phone.raw_inputs", None)
|
|
|
|
features[f"{ACTION}.enabled"] = PolicyFeature(type=FeatureType.ACTION, shape=(1,))
|
|
features[f"{ACTION}.target_x"] = PolicyFeature(type=FeatureType.ACTION, shape=(1,))
|
|
features[f"{ACTION}.target_y"] = PolicyFeature(type=FeatureType.ACTION, shape=(1,))
|
|
features[f"{ACTION}.target_z"] = PolicyFeature(type=FeatureType.ACTION, shape=(1,))
|
|
features[f"{ACTION}.target_wx"] = PolicyFeature(type=FeatureType.ACTION, shape=(1,))
|
|
features[f"{ACTION}.target_wy"] = PolicyFeature(type=FeatureType.ACTION, shape=(1,))
|
|
features[f"{ACTION}.target_wz"] = PolicyFeature(type=FeatureType.ACTION, shape=(1,))
|
|
features[f"{ACTION}.gripper"] = PolicyFeature(type=FeatureType.ACTION, shape=(1,))
|
|
return features
|