Aloha Physical Deployment#

Overview#

This document describes the complete workflow for deploying the FluxVLA model on the AgileX Aloha dual-arm platform for autonomous manipulation.

Power-On#

Turn on the robot power supply and power on all four robotic arms.
Verify that all four robotic arms display green indicator lights, confirming successful power-on.

2. Software Environment Preparation#

2.1 Create a Conda Environment#

conda create -n aloha_sdk3.0 python=3.8.19
conda activate aloha_sdk3.0

2.2 Install Piper SDK#

git clone https://github.com/agilexrobotics/piper_sdk.git
cd piper_sdk
pip3 install .

3. Launch the Robotic Arms#

Launch the robotic arm enable and control nodes (mode:=1 for follower mode, auto_enable:=true for automatic enabling):

conda activate aloha_sdk3.0
roslaunch piper start_ms_piper.launch mode:=1 auto_enable:=true

4. Launch the Cameras#

Launch the three depth cameras:

roslaunch astra_camera nmulti_camera.launch

After launching, verify that the following topics are being published correctly using rostopic list:

Camera	Color Topic	Depth Topic	Point Cloud Topic
Front (f)	`/camera_f/color/image_raw`	`/camera_f/depth/image_raw`	`/camera_f/depth/points`
Hand (h)	`/camera_h/color/image_raw`	`/camera_h/depth/image_raw`	`/camera_h/depth/points`
Left (l)	`/camera_l/color/image_raw`	`/camera_l/depth/image_raw`	`/camera_l/depth/points`
Right (r)	`/camera_r/color/image_raw`	`/camera_r/depth/image_raw`	`/camera_r/depth/points`

Robotic arm joint state topics:

Topic	Description
`/master/joint_left`	Left master arm joint states
`/master/joint_right`	Right master arm joint states
`/puppet/joint_left`	Left follower arm joint states
`/puppet/joint_right`	Right follower arm joint states
`/puppet/end_pose_left`	Left follower arm end-effector pose
`/puppet/end_pose_right`	Right follower arm end-effector pose
`/left_arm/trajectory`	Left arm trajectory command
`/right_arm/trajectory`	Right arm trajectory command
`/left_arm/model_trajectory`	Left arm model output trajectory
`/right_arm/model_trajectory`	Right arm model output trajectory

5. Run FluxVLA Model Inference#

Launch via “Run and Debug” in the IDE, or run directly from the command line:

WANDB_MODE=disabled HF_ENDPOINT=https://hf-mirror.com \
python scripts/inference_real_robot.py \
    --config configs/gr00t/gr00t_eagle_3b_aloha_fold_towel_3cam_4090_full_train.py \
    --ckpt-path ./work_dirs/work_dirs/419acaf40_gr00t_eagle_3b_aloha_fold_towel_3cam_4090_full_train_bs256/checkpoints/step-037356-epoch-06-loss=0.0617.pt

Corresponding VSCode launch.json configuration:

{
    "name": "Inference gr00t eagle",
    "type": "debugpy",
    "request": "launch",
    "program": "scripts/inference_real_robot.py",
    "console": "integratedTerminal",
    "args": [
        "--config", "configs/gr00t/gr00t_eagle_3b_aloha_fold_towel_3cam_4090_full_train.py",
        "--ckpt-path", "./work_dirs/work_dirs/419acaf40_gr00t_eagle_3b_aloha_fold_towel_3cam_4090_full_train_bs256/checkpoints/step-037356-epoch-06-loss=0.0617.pt"
    ],
    "env": {
        "WANDB_MODE": "disabled",
        "HF_ENDPOINT": "https://hf-mirror.com"
    }
}

6. Robotic Arm Reset#

Upon task completion or when a reset is required, send the reset command to return both arms to the zero position (with the end-effector gripper maintaining a slight opening of 0.1):

Left arm reset:

rostopic pub /master/joint_left sensor_msgs/JointState "header:
  seq: 0
  stamp: {secs: 0, nsecs: 0}
  frame_id: ''
name: ['']
position: [0,0,0,0,0,0,0.1]
velocity: [0]
effort: [0]"

Right arm reset:

rostopic pub /master/joint_right sensor_msgs/JointState "header:
  seq: 0
  stamp: {secs: 0, nsecs: 0}
  frame_id: ''
name: ['']
position: [0,0,0,0,0,0,0.1]
velocity: [0]
effort: [0]"

The 7 values in position correspond to 6 joint angles and 1 gripper opening value. Setting all to zero returns the arm to its initial pose, while 0.1 keeps the gripper in a slightly open state.