---
id: wiki-2026-0508-degrees-of-freedom
title: Degrees of Freedom (DOF)
category: 10_Wiki/Topics
status: verified
canonical_id: self
aliases: [DOF, degrees of freedom, kinematic DOF, statistical DOF]
duplicate_of: none
source_trust_level: A
confidence_score: 0.94
verification_status: applied
tags: [robotics, kinematics, statistics, mathematics, physics, mechanical]
raw_sources: []
last_reinforced: 2026-05-10
github_commit: pending
tech_stack:
  language: Math / Robotics / Statistics
  applicable_to: [Robotics, Statistics, Mechanics, ML]
---

# Degrees of Freedom (DOF)

## 매 한 줄
> **"매 system 의 independent parameter 의 count"**. 매 robotics: 매 movable joint. 매 statistics: 매 sample 의 free 의 vary. 매 modern AI: 매 model parameter (over-parameterization).

## 매 핵심

### 매 mechanical / robotics
- **Rigid body in 3D**: 6 DOF (3 translation + 3 rotation).
- **Rigid body in 2D**: 3 DOF.
- **Manipulator**: 매 joint 의 sum.
- **Mobile robot**: 매 base + arm.

### 매 Grübler-Kutzbach formula
```
DOF = λ(n - j - 1) + Σ f_i
```
- λ: 매 6 (3D), 3 (2D).
- n: 매 link 수 (base 포함).
- j: 매 joint 수.
- f_i: 매 joint 의 freedom.

### 매 statistical DOF
- **Sample variance**: n - 1 (mean 의 1 의 lose).
- **Linear regression**: n - p - 1 (p predictor).
- **Chi-square**: 매 categories - 1.
- **t-distribution**: 매 sample-dependent.

### 매 ML / DL
- **Effective DOF**: 매 model 의 parameter count.
- **Over-parameterization**: 매 modern DL 의 paradox (n_param >> n_data).
- **Implicit regularization**: 매 SGD 의 effective DOF 의 reduce.

### 매 응용
1. **Robot arm design**: 매 6+ DOF 의 redundancy.
2. **VR head tracking**: 6 DOF.
3. **Hand tracking**: 매 26 DOF (each hand).
4. **Statistics**: 매 hypothesis test.
5. **Camera pose**: 6 DOF.

## 💻 패턴

### Compute DOF (manipulator)
```python
def grubler_dof_3d(n_links, joints):
    """매 3D Grübler. joints: list of (type, freedom)."""
    j = len(joints)
    f_sum = sum(f for _, f in joints)
    return 6 * (n_links - j - 1) + f_sum

# 매 6-axis arm: 6 revolute joints
arm = grubler_dof_3d(7, [('rev', 1)] * 6)  # 6
# 매 4-bar linkage (planar)
def grubler_dof_2d(n, joints):
    return 3 * (n - len(joints) - 1) + sum(f for _, f in joints)
fourbar = grubler_dof_2d(4, [('rev', 1)] * 4)  # 1
```

### Statistical DOF (chi-square test)
```python
import scipy.stats as st

def chi2_dof(observed, expected):
    chi2 = sum((o - e)**2 / e for o, e in zip(observed, expected))
    dof = len(observed) - 1
    p = 1 - st.chi2.cdf(chi2, dof)
    return chi2, dof, p
```

### Sample variance (Bessel's correction)
```python
import numpy as np

def variance(data, ddof=1):  # 매 ddof = degrees of freedom adjust
    n = len(data)
    mean = sum(data) / n
    return sum((x - mean)**2 for x in data) / (n - ddof)

# 매 ddof=1: 매 sample (unbiased)
# 매 ddof=0: 매 population
```

### Effective DOF (ridge regression)
```python
def effective_dof_ridge(X, lam):
    """매 trace of hat matrix."""
    XtX = X.T @ X
    n_features = X.shape[1]
    H = X @ np.linalg.inv(XtX + lam * np.eye(n_features)) @ X.T
    return np.trace(H)
```

### 6-DOF pose (3D vision)
```python
class Pose6DOF:
    def __init__(self, position, orientation):
        self.t = np.array(position)  # 매 [x, y, z]
        self.q = orientation  # 매 quaternion [w, x, y, z]
    
    def as_matrix(self):
        T = np.eye(4)
        T[:3, 3] = self.t
        # 매 quaternion → rotation matrix (Hamilton)
        w, x, y, z = self.q
        T[:3, :3] = np.array([
            [1-2*(y*y+z*z), 2*(x*y-z*w), 2*(x*z+y*w)],
            [2*(x*y+z*w), 1-2*(x*x+z*z), 2*(y*z-x*w)],
            [2*(x*z-y*w), 2*(y*z+x*w), 1-2*(x*x+y*y)],
        ])
        return T
```

### Hand tracking DOF
```python
HAND_DOF = {
    'wrist': 6,         # 매 free in space
    'thumb': 5,         # 매 CMC 의 2, MCP 의 1, IP 의 1, opposition 의 1
    'index_finger': 4,  # 매 MCP 2 + PIP 1 + DIP 1
    'middle_finger': 4,
    'ring_finger': 4,
    'pinky_finger': 4,
}
total = sum(HAND_DOF.values())  # 27
```

### Redundancy (kinematic)
```python
def redundancy(robot_dof, task_dof):
    """매 redundancy = robot_dof - task_dof."""
    return max(0, robot_dof - task_dof)

# 매 7-DOF arm + 매 6-DOF task → 1-DOF redundancy
# 매 self-motion 의 obstacle 의 avoid
print(redundancy(7, 6))  # 1
```

### Over-parameterization (DL)
```python
def model_dof(model):
    """매 deep learning effective DOF approximation."""
    total = sum(p.numel() for p in model.parameters())
    trainable = sum(p.numel() for p in model.parameters() if p.requires_grad)
    return {'total': total, 'trainable': trainable}

# 매 GPT-3: 175B parameters >> 매 dataset size
# 매 overfitting paradox: 매 implicit regularization 의 explain
```

## 매 결정 기준
| 상황 | DOF Approach |
|---|---|
| Manipulator design | Grübler-Kutzbach |
| Robot redundancy | n_robot - n_task |
| Statistics test | Specific test formula |
| ML model | Parameter count + effective |
| VR / AR | 6 DOF (full) or 3 DOF (rotation) |
| Hand tracking | ~26 per hand |

**기본값**: 매 task-specific DOF + 매 redundancy 의 prefer (manipulator) + 매 statistical 의 ddof aware.

## 🔗 Graph
- 부모: [[Mechanics]] · [[Statistics]] · [[Robotics]]
- 변형: [[Kinematic-DOF]] · [[Statistical-DOF]] · [[Effective-DOF]]
- 응용: [[Denavit-Hartenberg-Parameters]] · [[Manipulator]] · [[Hypothesis-Testing]]
- Adjacent: [[Over-Parameterization]] · [[Bessel-Correction]] · [[Pose-Estimation]]

## 🤖 LLM 활용
**언제**: 매 robot design. 매 statistical analysis. 매 ML capacity.
**언제 X**: 매 informal usage.

## ❌ 안티패턴
- **Confuse statistical / mechanical**: 매 different concept.
- **Forget Bessel correction**: 매 biased estimator.
- **Under-DOF manipulator**: 매 task 의 reach X.
- **Over-DOF without redundancy logic**: 매 self-collision.

## 🧪 검증 / 중복
- Verified (Spong Robot Dynamics, Statistics textbook).
- 신뢰도 A.

## 🕓 Changelog
| 날짜 | 변경 |
|---|---|
| 2026-04-20 | Auto-reinforced |
| 2026-05-08 | Phase 1 |
| 2026-05-10 | Manual cleanup — mechanical + statistical DOF + 매 Grübler / chi2 / pose / redundancy code |