Hugging Face's logo

Datasets:
farouqh
/
mesh-corpus-v1

Tasks:
Robotics
Reinforcement Learning
Languages:
English
Size:
1K<n<10K
Tags:
robotics
synthetic-data
mujoco
navigation
mobile-robot
simulation
License:
Dataset card Files
xet
 Community
Dataset Viewer
Duplicate
The dataset viewer is not available for this split.
Cannot load the dataset split (in streaming mode) to extract the first rows.
Error code:   StreamingRowsError
Exception:    CastError
Message:      Couldn't cast
version: string
description: string
episode_fields: struct<episode_index: struct<type: string, desc: string>, env_id: struct<type: string, desc: string> (... 825 chars omitted)
  child 0, episode_index: struct<type: string, desc: string>
      child 0, type: string
      child 1, desc: string
  child 1, env_id: struct<type: string, desc: string>
      child 0, type: string
      child 1, desc: string
  child 2, env_category: struct<type: string, desc: string>
      child 0, type: string
      child 1, desc: string
  child 3, seed: struct<type: string, desc: string>
      child 0, type: string
      child 1, desc: string
  child 4, physics_preset: struct<type: string, desc: string>
      child 0, type: string
      child 1, desc: string
  child 5, success: struct<type: string, desc: string>
      child 0, type: string
      child 1, desc: string
  child 6, failure_reason: struct<type: string, desc: string>
      child 0, type: string
      child 1, desc: string
  child 7, duration_s: struct<type: string, desc: string>
      child 0, type: string
      child 1, desc: string
  child 8, path_length_m: struct<type: string, desc: string>
      child 0, type: string
      child 1, desc: string
  child 9, energy_joules: struct<type: string, desc: string>
      child 0, type: string
      child 1, desc: string
  child 10, collision_count: struct<type: string, desc: string>
      child 0, type: string
      child 1, desc: string
  child 11, collision_occurred: struct<type: string, desc
...
   child 1, pos: list<item: double>
          child 0, item: double
      child 2, quat: list<item: double>
          child 0, item: double
      child 3, linvel: list<item: double>
          child 0, item: double
      child 4, angvel: list<item: double>
          child 0, item: double
      child 5, touch: list<item: double>
          child 0, item: double
      child 6, imu_accel: list<item: double>
          child 0, item: double
      child 7, imu_gyro: list<item: double>
          child 0, item: double
      child 8, wheel_vel: list<item: double>
          child 0, item: double
      child 9, ctrl: list<item: double>
          child 0, item: double
      child 10, contact_force: double
      child 11, n_contacts: int64
      child 12, dist_to_goal: double
summary: struct<episode_index: int64, env_id: string, env_category: string, seed: int64, physics_preset: stri (... 313 chars omitted)
  child 0, episode_index: int64
  child 1, env_id: string
  child 2, env_category: string
  child 3, seed: int64
  child 4, physics_preset: string
  child 5, success: bool
  child 6, failure_reason: string
  child 7, duration_s: double
  child 8, path_length_m: double
  child 9, energy_joules: double
  child 10, collision_count: int64
  child 11, collision_occurred: bool
  child 12, n_traj_frames: int64
  child 13, floor_friction: double
  child 14, robot_mass_scale: double
  child 15, obstacle_mass_scale: double
  child 16, actuator_noise_std: double
  child 17, sensor_noise_std: double
to
{'summary': {'episode_index': Value('int64'), 'env_id': Value('string'), 'env_category': Value('string'), 'seed': Value('int64'), 'physics_preset': Value('string'), 'success': Value('bool'), 'failure_reason': Value('string'), 'duration_s': Value('float64'), 'path_length_m': Value('float64'), 'energy_joules': Value('float64'), 'collision_count': Value('int64'), 'collision_occurred': Value('bool'), 'n_traj_frames': Value('int64'), 'floor_friction': Value('float64'), 'robot_mass_scale': Value('float64'), 'obstacle_mass_scale': Value('float64'), 'actuator_noise_std': Value('float64'), 'sensor_noise_std': Value('float64')}, 'trajectory': List({'t': Value('float64'), 'pos': List(Value('float64')), 'quat': List(Value('float64')), 'linvel': List(Value('float64')), 'angvel': List(Value('float64')), 'touch': List(Value('float64')), 'imu_accel': List(Value('float64')), 'imu_gyro': List(Value('float64')), 'wheel_vel': List(Value('float64')), 'ctrl': List(Value('float64')), 'contact_force': Value('float64'), 'n_contacts': Value('int64'), 'dist_to_goal': Value('float64')})}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/utils.py", line 149, in get_rows_or_raise
                  return get_rows(
                      dataset=dataset,
                  ...<4 lines>...
                      column_names=column_names,
                  )
                File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
                  return func(*args, **kwargs)
                File "/src/services/worker/src/worker/utils.py", line 129, in get_rows
                  rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
                File "/src/services/worker/src/worker/utils.py", line 489, in safe_iter
                  yield from ds.decode(False) if ds.features else ds
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2818, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2355, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2380, in _iter_arrow
                  for key, pa_table in self.ex_iterable._iter_arrow():
                                       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 536, in _iter_arrow
                  for key, pa_table in iterator:
                                       ^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 419, in _iter_arrow
                  for key, pa_table in self.generate_tables_fn(**gen_kwags):
                                       ~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 343, in _generate_tables
                  self._cast_table(pa_table, json_field_paths=json_field_paths),
                  ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
                  pa_table = table_cast(pa_table, self.info.features.arrow_schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2369, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2297, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              version: string
              description: string
              episode_fields: struct<episode_index: struct<type: string, desc: string>, env_id: struct<type: string, desc: string> (... 825 chars omitted)
                child 0, episode_index: struct<type: string, desc: string>
                    child 0, type: string
                    child 1, desc: string
                child 1, env_id: struct<type: string, desc: string>
                    child 0, type: string
                    child 1, desc: string
                child 2, env_category: struct<type: string, desc: string>
                    child 0, type: string
                    child 1, desc: string
                child 3, seed: struct<type: string, desc: string>
                    child 0, type: string
                    child 1, desc: string
                child 4, physics_preset: struct<type: string, desc: string>
                    child 0, type: string
                    child 1, desc: string
                child 5, success: struct<type: string, desc: string>
                    child 0, type: string
                    child 1, desc: string
                child 6, failure_reason: struct<type: string, desc: string>
                    child 0, type: string
                    child 1, desc: string
                child 7, duration_s: struct<type: string, desc: string>
                    child 0, type: string
                    child 1, desc: string
                child 8, path_length_m: struct<type: string, desc: string>
                    child 0, type: string
                    child 1, desc: string
                child 9, energy_joules: struct<type: string, desc: string>
                    child 0, type: string
                    child 1, desc: string
                child 10, collision_count: struct<type: string, desc: string>
                    child 0, type: string
                    child 1, desc: string
                child 11, collision_occurred: struct<type: string, desc
              ...
                 child 1, pos: list<item: double>
                        child 0, item: double
                    child 2, quat: list<item: double>
                        child 0, item: double
                    child 3, linvel: list<item: double>
                        child 0, item: double
                    child 4, angvel: list<item: double>
                        child 0, item: double
                    child 5, touch: list<item: double>
                        child 0, item: double
                    child 6, imu_accel: list<item: double>
                        child 0, item: double
                    child 7, imu_gyro: list<item: double>
                        child 0, item: double
                    child 8, wheel_vel: list<item: double>
                        child 0, item: double
                    child 9, ctrl: list<item: double>
                        child 0, item: double
                    child 10, contact_force: double
                    child 11, n_contacts: int64
                    child 12, dist_to_goal: double
              summary: struct<episode_index: int64, env_id: string, env_category: string, seed: int64, physics_preset: stri (... 313 chars omitted)
                child 0, episode_index: int64
                child 1, env_id: string
                child 2, env_category: string
                child 3, seed: int64
                child 4, physics_preset: string
                child 5, success: bool
                child 6, failure_reason: string
                child 7, duration_s: double
                child 8, path_length_m: double
                child 9, energy_joules: double
                child 10, collision_count: int64
                child 11, collision_occurred: bool
                child 12, n_traj_frames: int64
                child 13, floor_friction: double
                child 14, robot_mass_scale: double
                child 15, obstacle_mass_scale: double
                child 16, actuator_noise_std: double
                child 17, sensor_noise_std: double
              to
              {'summary': {'episode_index': Value('int64'), 'env_id': Value('string'), 'env_category': Value('string'), 'seed': Value('int64'), 'physics_preset': Value('string'), 'success': Value('bool'), 'failure_reason': Value('string'), 'duration_s': Value('float64'), 'path_length_m': Value('float64'), 'energy_joules': Value('float64'), 'collision_count': Value('int64'), 'collision_occurred': Value('bool'), 'n_traj_frames': Value('int64'), 'floor_friction': Value('float64'), 'robot_mass_scale': Value('float64'), 'obstacle_mass_scale': Value('float64'), 'actuator_noise_std': Value('float64'), 'sensor_noise_std': Value('float64')}, 'trajectory': List({'t': Value('float64'), 'pos': List(Value('float64')), 'quat': List(Value('float64')), 'linvel': List(Value('float64')), 'angvel': List(Value('float64')), 'touch': List(Value('float64')), 'imu_accel': List(Value('float64')), 'imu_gyro': List(Value('float64')), 'wheel_vel': List(Value('float64')), 'ctrl': List(Value('float64')), 'contact_force': Value('float64'), 'n_contacts': Value('int64'), 'dist_to_goal': Value('float64')})}
              because column names don't match

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

Mesh Corpus v1 — MuJoCo Warehouse Navigation Dataset

Physically accurate synthetic training data for mobile robot navigation policies.

Every sensor reading in this dataset comes from MuJoCo's constraint solver and forward kinematics — not from random number generators. This makes it suitable for training real robot policies where causal consistency between sensor modalities matters.

Dataset Summary

Property Value
Physics engine MuJoCo 3.x (headless)
Robot type Mobile holonomic chassis (freejoint)
Task Point-to-point navigation in warehouse environments
Environments 10 procedurally generated warehouse layouts
Episodes (sample) 10 (full 100-episode corpus available via pipeline)
Success rate ~40% (realistic collision failures)
Sensor modalities 5 (position, orientation, IMU, wheel odometry, touch)
Trajectory frequency ~100 Hz (one frame per 5 sim steps at DT=0.002s)
Schema version 2.0.0
License MIT (free for academic use, commercial license available)

What Makes This Different From LLM-Generated Data

An LLM can generate numbers that look like robot sensor data. MuJoCo generates numbers that are physics.

When the robot hits a wall in this dataset:

  • contact_force spikes at the exact collision timestep
  • touch bumper registers the correct Newton value
  • linvel drops to zero
  • pos stops updating
  • failure_reason is set to "collision"

All of these happen simultaneously, causally linked by the physics solver. No language model can reproduce this consistency without running the simulation.

Sensor Schema (per trajectory frame)

Field Shape Description
t scalar Simulation time (s)
pos [3] World position [x, y, z] (m)
quat [4] Orientation quaternion [w, x, y, z]
linvel [3] Linear velocity (m/s)
angvel [3] Angular velocity (rad/s)
touch [4] Bumper touch sensors [front, rear, left, right] (N)
imu_accel [3] IMU accelerometer (m/s²)
imu_gyro [3] IMU gyroscope (rad/s)
wheel_vel [12] Wheel site velocities [fl, fr, rl, rr × xyz] (m/s)
ctrl [3] Control signal [vx, vy, omega]
contact_force scalar Max contact force magnitude (N)
n_contacts scalar Number of active contacts
dist_to_goal scalar Distance to goal (m)

Episode Summary Fields

Each episode includes a summary with:

  • success (bool), failure_reason (collision / timeout / null)
  • duration_s, path_length_m, energy_joules, collision_count
  • floor_friction, robot_mass_scale (physics randomisation params)
  • env_id, env_category (standard / edge_case / dynamic)

Files

File Description
sample_10episodes.jsonl 10 complete episodes with full trajectory data (~11 MB)
summaries.json Episode metadata for all 100 episodes (no trajectory)
schema.json Full field definitions and types
splits.json Train/val/test split indices
mesh_corpus_v2.h5 HDF5 format — compressed sensor arrays (20 MB, 100 episodes)

Reproduce or Scale

The full pipeline is open source. Generate your own corpus:

git clone https://github.com/farouqh/mesh-twin
cd mesh-twin
pip install -r requirements.txt

# Generate 500 episodes across 30 environments
python -m mesh_twin.dataset_packager \
  --run-pipeline \
  --robot tests/fixtures/mobile_base.urdf \
  --episodes 500 --seed 42 --n-envs 30 \
  --output dataset/

Commercial Licensing

This dataset is free for academic and non-commercial research use (MIT license).

For commercial use, custom corpus generation (your robot, your environment, your task), or enterprise licensing, contact: farouqsultan@gmail.com

Citation 

@dataset{mesh_corpus_v1_2025,
  title     = {Mesh Corpus v1: MuJoCo Warehouse Navigation Dataset},
  author    = {Sultan, Farouq},
  year      = {2025},
  publisher = {Hugging Face},
  url       = {https://huggingface.co/datasets/farouqh/mesh-corpus-v1},
  note      = {Physically accurate synthetic training data for mobile robot navigation policies}
}

Related Work

Downloads last month
43
Total file size:
32.1 MB