Data Recording

This section covers the data recording capabilities of the DexRobot MuJoCo node.

Recording Options

The node supports three recording formats:

  1. CSV files (tracked states and sensors)

  2. ROS bags (all ROS topics)

  3. MP4 videos (simulation visualization)

These can be used individually or simultaneously.

CSV Recording

Configuration

Enable CSV recording:

python nodes/dexrobot_mujoco_ros.py model.xml \
    --output-formats csv \
    --output-csv-path data.csv

Data Format

CSV columns include:

  • timestamp: Nanosecond timestamp

  • {joint_name}_pos: Joint positions

  • {joint_name}_vel: Joint velocities

  • {body_name}_pos: Body positions (x, y, z)

  • {body_name}_quat: Body orientations (w, x, y, z)

  • {sensor_name}: Sensor readings

Example CSV:

timestamp,r_f_joint1_1_pos,r_f_joint1_1_vel,...
1234567890,0.5,0.1,...
1234567900,0.51,0.1,...

ROS Bag Recording

Configuration

Enable bag recording:

python nodes/dexrobot_mujoco_ros.py model.xml \
    --output-formats ros \
    --output-bag-path recording.bag \
    --additional-bag-topics /camera/image_raw

Recorded Topics

Default topics:

  • /joint_commands

  • /joint_states

  • /body_poses

  • /touch_sensors

Additional topics can be specified with --additional-bag-topics.

MP4 Recording

Configuration

Enable MP4 recording:

python nodes/dexrobot_mujoco_ros.py model.xml \
    --output-formats mp4 \
    --output-mp4-path video.mp4 \
    --renderer-dimension 640,480

Parameters:

  • Frame rate: 20 FPS (fixed)

  • Codec: mp4v

  • Resolution: Specified by renderer-dimension

Combined Recording

Enable all formats:

python nodes/dexrobot_mujoco_ros.py model.xml \
    --output-formats ros csv mp4 \
    --output-csv-path data.csv \
    --output-bag-path recording.bag \
    --output-mp4-path video.mp4

Data Processing

CSV Processing

Example Python script:

import pandas as pd
import numpy as np

def analyze_data(csv_path):
    # Load data
    data = pd.read_csv(csv_path)

    # Process joint positions
    joint_cols = [col for col in data.columns
                 if col.endswith('_pos')]
    joint_data = data[joint_cols]

    # Calculate statistics
    stats = {
        'mean': joint_data.mean(),
        'std': joint_data.std(),
        'max': joint_data.max(),
        'min': joint_data.min()
    }

    # Extract touch data
    touch_cols = [col for col in data.columns
                 if 'touch' in col]
    touch_data = data[touch_cols]

    return stats, touch_data

ROS Bag Processing

Example Python script:

import rosbag
from sensor_msgs.msg import JointState

def process_bag(bag_path):
    joint_data = []

    with rosbag.Bag(bag_path) as bag:
        # Process joint states
        for topic, msg, t in bag.read_messages(
            topics=['joint_states']
        ):
            joint_data.append({
                'time': t.to_sec(),
                'positions': msg.position,
                'velocities': msg.velocity
            })

    return joint_data

Example Pipeline

Complete Recording Setup

  1. Create configuration:

    # config/recording.yaml
camera:
  distance: 1.5
  elevation: -20
  azimuth: 0
  lookat: [0.0, 0.0, 1.0]

tracked_joints:
  - [r_f_joint1_1, r_f_joint1_2]

tracked_bodies:
  - [right_hand_base]

tracked_sensors:
  - [touch_r_f_link1_4]

  2. Launch recording:

    python nodes/dexrobot_mujoco_ros.py model.xml \
    --config config/recording.yaml \
    --output-formats ros csv mp4 \
    --output-csv-path data/trial1.csv \
    --output-bag-path data/trial1.bag \
    --output-mp4-path data/trial1.mp4

  3. Process data:

    from pathlib import Path
import pandas as pd
import rosbag

def process_trial(trial_dir):
    # Load CSV
    csv_data = pd.read_csv(
        trial_dir / 'trial1.csv'
    )

    # Process bag
    bag_data = []
    with rosbag.Bag(trial_dir / 'trial1.bag') as bag:
        for topic, msg, t in bag.read_messages():
            bag_data.append((topic, t, msg))

    # Analyze data
    results = analyze_data(csv_data, bag_data)

    return results

Next Steps