Update README.md

README.md

@@ -3,48 +3,43 @@
 Official release of the dataset from the paper:  
 **[Neural Motion Simulator: Pushing the Limit of World Models in Reinforcement Learning](https://arxiv.org/pdf/2504.07095)**  
 
-This dataset provides the training and evaluation data used for the **MoSim (Neural Motion Simulator)** world models.  
-All trajectories are collected **exclusively using random policies** across multiple standard simulation environments.  
+This dataset contains sequential state-action trajectories for **MoSim (Neural Motion Simulator)** world model training.  
+All trajectories are collected **from random policies** in classical control and locomotion environments.
 
 ---
 
 ## 📦 Dataset Overview
 
-- **Format**: Parquet files for efficient loading and processing.  
+- **Format**: `.npz` (NumPy compressed arrays)  
 - **Contents**:  
-  - State sequences (`qpos`, `qvel`, sensor readings)  
-  - Action sequences (raw control inputs)  
-  - Environment metadata (episode length, environment name)  
-- **Size**: ~XX GB (after compression)  
-- **Environments**: Includes standard MuJoCo-based tasks such as `Cheetah`, `Hopper`, `Humanoid`, and `Go2`.  
+  - `*_random.npz`: training episodes  
+  - `*_random_test.npz`: test episodes  
+- **Episode length**: 1000 steps per episode  
 
 ---
 
-## 📊 Structure
+## 📊 Data Structure
 
-Each trajectory is stored as a sequence of steps:
+Each `.npz` file contains:
 
-| Field          | Shape (per step)       | Description                               |
-|----------------|-----------------------|-------------------------------------------|
-| `qpos`         | *(n_joints,)*         | Joint positions                            |
-| `qvel`         | *(n_joints,)*         | Joint velocities                           |
-| `action`       | *(action_dim,)*       | Environment action                         |
-| `reward`       | *1*                   | Immediate reward                           |
-| `done`         | *1*                   | Episode termination flag                   |
-| `env_id`       | *string*              | Environment identifier                     |
+| Key        | Shape                               | Description                        |
+|------------|-------------------------------------|------------------------------------|
+| `states`   | *(num_episodes, 1000, state_dim)*    | State at each timestep             |
+| `actions`  | *(num_episodes, 1000, action_dim)*   | Action applied at each timestep    |
 
-All episodes are segmented and zero-padded to fixed length for training convenience.
+**State composition**:
 
----
-
-## 📥 Usage
+1. **Joint DOF (articulated body)**  
+   - Joint angles *(radians)*  
+   - Joint angular velocities *(rad/s)*  
+2. **Root DOF (global free body)**  
+   - Root position *(x, y, z)*  
+   - Root linear velocity *(vx, vy, vz)*  
+3. **Root Orientation & Rotation**  
+   - Root rotation quaternion *(qx, qy, qz, qw)*  
+   - Root angular velocity *(wx, wy, wz)*  
 
-```python
-from datasets import load_dataset
+> ⚡ For manipulation control tasks like `Panda`, only joint angles and velocities are provided.  
+> ⚡ For locomotion tasks like `Humanoid` or `Go2`, full root DOF and velocities are included.
 
-# Load the full MoSim dataset
-dataset = load_dataset("wujiss1/MoSim_Dataset")
 
-# Access one environment subset
-hopper_data = dataset["train"].filter(lambda x: x["env_id"] == "Hopper")
-print(hopper_data[0].keys())