Update Documentation For Aloha Solo

docs/files/act_aloha_solo_real.yaml

@@ -0,0 +1,117 @@
+# @package _global_
+
+# Use `act_aloha_solo_real.yaml` to train on real-world datasets collected on Aloha or Aloha-2 robots.
+# It contains 2 cameras (i.e. cam_right_wrist/cam_left_wrist, cam_high).
+# Also, `training.eval_freq` is set to -1. This config is used to evaluate checkpoints at a certain frequency of training steps.
+# When it is set to -1, it deactivates evaluation. This is because real-world evaluation is done through our `control_robot.py` script.
+# Look at the documentation in header of `control_robot.py` for more information on how to collect data , train and evaluate a policy.
+#
+# Example of usage for training and inference with `control_robot.py`:
+# ```bash
+# python lerobot/scripts/train.py \
+#   policy=act_aloha_solo_real \
+#   env=aloha_solo_real
+# ```
+#
+# Example of usage for training and inference with [Dora-rs](https://github.com/dora-rs/dora-lerobot):
+# ```bash
+# python lerobot/scripts/train.py \
+#   policy=act_aloha_solo_real \
+#   env=dora_aloha_solo_real
+# ```
+
+seed: 1000
+dataset_repo_id: lerobot/aloha_solo_lego
+
+override_dataset_stats:
+  observation.images.cam_right_wrist:
+    # stats from imagenet, since we use a pretrained vision model
+    mean: [[[0.485]], [[0.456]], [[0.406]]]  # (c,1,1)
+    std: [[[0.229]], [[0.224]], [[0.225]]]  # (c,1,1)
+  # Uncomment for left wrist and comment the right wrist
+  # observation.images.cam_left_wrist:
+  #   # stats from imagenet, since we use a pretrained vision model
+  #   mean: [[[0.485]], [[0.456]], [[0.406]]]  # (c,1,1)
+  #   std: [[[0.229]], [[0.224]], [[0.225]]]  # (c,1,1)
+  observation.images.cam_high:
+    # stats from imagenet, since we use a pretrained vision model
+    mean: [[[0.485]], [[0.456]], [[0.406]]]  # (c,1,1)
+    std: [[[0.229]], [[0.224]], [[0.225]]]  # (c,1,1)
+
+
+training:
+  offline_steps: 80000
+  online_steps: 0
+  eval_freq: -1
+  save_freq: 10000
+  log_freq: 100
+  save_checkpoint: true
+
+  batch_size: 8
+  lr: 1e-5
+  lr_backbone: 1e-5
+  weight_decay: 1e-4
+  grad_clip_norm: 10
+  online_steps_between_rollouts: 1
+
+  delta_timestamps:
+    action: "[i / ${fps} for i in range(${policy.chunk_size})]"
+
+eval:
+  n_episodes: 50
+  batch_size: 50
+
+# See `configuration_act.py` for more details.
+policy:
+  name: act
+
+  # Input / output structure.
+  n_obs_steps: 1
+  chunk_size: 100
+  n_action_steps: 100
+
+  input_shapes:
+    # TODO(rcadene, alexander-soare): add variables for height and width from the dataset/env?
+    observation.images.cam_right_wrist: [3, 480, 640]
+    # observation.images.cam_left_wrist: [3, 480, 640] # Uncomment if using left wrist and comment the right.
+    observation.images.cam_high: [3, 480, 640]
+    observation.state: ["${env.state_dim}"]
+  output_shapes:
+    action: ["${env.action_dim}"]
+
+  # Normalization / Unnormalization
+  input_normalization_modes:
+    observation.images.cam_right_wrist: mean_std
+    # observation.images.cam_left_wrist: mean_std # Uncomment if using left wrist and comment the right.
+    observation.images.cam_high: mean_std
+    observation.state: mean_std
+  output_normalization_modes:
+    action: mean_std
+
+  # Architecture.
+  # Vision backbone.
+  vision_backbone: resnet18
+  pretrained_backbone_weights: ResNet18_Weights.IMAGENET1K_V1
+  replace_final_stride_with_dilation: false
+  # Transformer layers.
+  pre_norm: false
+  dim_model: 512
+  n_heads: 8
+  dim_feedforward: 3200
+  feedforward_activation: relu
+  n_encoder_layers: 4
+  # Note: Although the original ACT implementation has 7 for `n_decoder_layers`, there is a bug in the code
+  # that means only the first layer is used. Here we match the original implementation by setting this to 1.
+  # See this issue https://github.com/tonyzhaozh/act/issues/25#issue-2258740521.
+  n_decoder_layers: 1
+  # VAE.
+  use_vae: true
+  latent_dim: 32
+  n_vae_encoder_layers: 4
+
+  # Inference.
+  temporal_ensemble_momentum: null
+
+  # Training and loss computation.
+  dropout: 0.1
+  kl_weight: 10.0

docs/files/aloha_solo.yaml

@@ -0,0 +1,78 @@
+# [Aloha: A Low-Cost Hardware for Bimanual Teleoperation](https://www.trossenrobotics.com/aloha-stationary)
+# https://aloha-2.github.io
+
+# Requires installing extras packages
+# With pip: `pip install -e ".[dynamixel intelrealsense]"`
+# With poetry: `poetry install --sync --extras "dynamixel intelrealsense"`
+
+# See [tutorial](https://github.com/huggingface/lerobot/blob/main/examples/9_use_aloha.md)
+
+
+_target_: lerobot.common.robot_devices.robots.manipulator.ManipulatorRobot
+robot_type: aloha
+# Specific to Aloha, LeRobot comes with default calibration files. Assuming the motors have been
+# properly assembled, no manual calibration step is expected. If you need to run manual calibration,
+# simply update this path to ".cache/calibration/aloha"
+calibration_dir: .cache/calibration/aloha_default
+
+# /!\ FOR SAFETY, READ THIS /!\
+# `max_relative_target` limits the magnitude of the relative positional target vector for safety purposes.
+# Set this to a positive scalar to have the same value for all motors, or a list that is the same length as
+# the number of motors in your follower arms.
+# For Aloha, for every goal position request, motor rotations are capped at 5 degrees by default.
+# When you feel more confident with teleoperation or running the policy, you can extend
+# this safety limit and even removing it by setting it to `null`.
+# Also, everything is expected to work safely out-of-the-box, but we highly advise to
+# first try to teleoperate the grippers only (by commenting out the rest of the motors in this yaml),
+# then to gradually add more motors (by uncommenting), until you can teleoperate both arms fully
+max_relative_target: null
+
+leader_arms:
+  right:
+    _target_: lerobot.common.robot_devices.motors.dynamixel.DynamixelMotorsBus
+    port: /dev/ttyDXL_leader_right
+    motors:  # window_x
+      # name: (index, model)
+      waist: [1, xm430-w350]
+      shoulder: [2, xm430-w350]
+      shoulder_shadow: [3, xm430-w350]
+      elbow: [4, xm430-w350]
+      elbow_shadow: [5, xm430-w350]
+      forearm_roll: [6, xm430-w350]
+      wrist_angle: [7, xm430-w350]
+      wrist_rotate: [8, xl430-w250]
+      gripper: [9, xc430-w150]
+
+follower_arms:
+  right:
+    _target_: lerobot.common.robot_devices.motors.dynamixel.DynamixelMotorsBus
+    port: /dev/ttyDXL_follower_right
+    motors:
+      # name: [index, model]
+      waist: [1, xm540-w270]
+      shoulder: [2, xm540-w270]
+      shoulder_shadow: [3, xm540-w270]
+      elbow: [4, xm540-w270]
+      elbow_shadow: [5, xm540-w270]
+      forearm_roll: [6, xm540-w270]
+      wrist_angle: [7, xm540-w270]
+      wrist_rotate: [8, xm430-w350]
+      gripper: [9, xm430-w350]
+
+# Troubleshooting: If one of your IntelRealSense cameras freeze during
+# data recording due to bandwidth limit, you might need to plug the camera
+# on another USB hub or PCIe card.
+cameras:
+  cam_high:
+    _target_: lerobot.common.robot_devices.cameras.intelrealsense.IntelRealSenseCamera
+    serial_number: 1234567890
+    fps: 30
+    width: 640
+    height: 480
+
+  cam_right_wrist:
+    _target_: lerobot.common.robot_devices.cameras.intelrealsense.IntelRealSenseCamera
+    serial_number: 1234567890
+    fps: 30
+    width: 640
+    height: 480

docs/files/aloha_solo_real.yaml

@@ -0,0 +1,10 @@
+# @package _global_
+
+fps: 30
+
+env:
+  name: real_world
+  task: null
+  state_dim: 9
+  action_dim: 9
+  fps: ${fps}

docs/getting_started.rst

@@ -15,3 +15,4 @@ Getting Started
 
   ./getting_started/mobile.rst
   ./getting_started/stationary.rst
+  ./getting_started/solo.rst

docs/getting_started/mobile/software_setup.rst

@@ -77,12 +77,24 @@ To build the SLATE base control software, run the following commands:
 ALOHA Software Installation
 ===========================
 
+.. admonition:: Important Compatibility Notice
+
+   This documentation and software correspond to **Aloha version 2.0**.
+   It supports features such as teleoperation, data recording, replay, and visualization. 
+   However, it is **not compatible** with training and evaluation for **ACT** or **ACT++**.
+
+   For end-to-end training of **Aloha Stationary** and **Aloha Mobile**, stick to **Aloha 1.0**. 
+   You can find the Aloha 1.0 documentation here: `Aloha 1.0 Documentation <https://docs.trossenrobotics.com/aloha_docs>`_
+
+   We are actively working on updates to provide full compatibility in the future.
+
+
 1.  Clone the Interbotix fork of ALOHA into the workspace's source directory:
 
   .. code-block:: bash
 
     $ cd ~/interbotix_ws/src
-    $ git clone https://github.com/Interbotix/aloha.git
+    $ git clone https://github.com/Interbotix/aloha.git -b 2.0
 
 2.  Run rosdep to install any dependencies:
 
@@ -95,13 +107,7 @@ ALOHA Software Installation
 
 .. _`link`: https://github.com/Interbotix/interbotix_ros_toolboxes/blob/c187bcea89b60391244bb19943ebd78f770aa975/interbotix_xs_toolbox/interbotix_xs_modules/interbotix_xs_modules/xs_robot/arm.py#L81
 
-4.  Add the following line to your ``~/.bashrc`` file:
-
-  .. code-block:: bash
-
-    $ echo "export INTERBOTIX_ALOHA_IS_MOBILE=true" >> ~/.bashrc
-
-5.  Build the workspace:
+4.  Build the workspace:
 
   .. code-block:: bash
 
@@ -199,7 +205,7 @@ Camera Setup
   .. image:: images/rsviewer_serialno.png
     :align: center
 
-4.  Put the camera serial number in the appropriate config entry at ``~/interbotix_ws/src/aloha/config/rs_cam.yaml``.
+4.  Put the camera serial number in the appropriate config entry at ``~/interbotix_ws/src/aloha/config/robot/aloha_mobile.yaml``.
 
 5.  Repeat for the rest of the cameras.
     If the workspace has not been symbolically-linked, a rebuild may be necessary.

docs/getting_started/solo.rst

@@ -0,0 +1,9 @@
+================================
+Solo ALOHA Getting Started
+================================
+
+.. toctree::
+  :maxdepth: 2
+
+  ./solo/hardware_setup.rst
+  ./solo/software_setup.rst

docs/getting_started/solo/hardware_setup.rst

@@ -0,0 +1,32 @@
+===========================
+Solo ALOHA Hardware Setup
+===========================
+
+.. attention::
+
+  This page is under construction as we work to provide images and more comprehensive videos.
+
+If you need any assistance or have questions about this process, please `contact our customer service`_.
+
+.. _`contact our customer service`: https://www.trossenrobotics.com/support
+
+Assembly Guide
+==============
+
+Video Overview
+--------------
+
+.. youtube:: jh21xCD_H0w
+    :align: center
+
+Steps
+-----
+
+
+#. Attach the camera mount and fix the camera securely on the Viper arm.
+#. Find a stable, flat surface to place your arms. Once satisfied, secure them using `Gorilla Tape <https://a.co/d/2v591vP>`_ or G-Clamps.
+#. Connect the power cables to the arms.
+#. Position the tripod so the camera effectively covers the experiment space.
+   **Warning:** Keep the camera position consistent throughout data collection and evaluation for optimal results.
+#. Connect the camera cables and arm cables to the USB hub.
+#. Connect the USB hub to your computer.