Humanoid Degrees of Freedom: Arm, Hand, and Leg Specs Compared

Understanding Degrees of Freedom in Humanoid Robotics

In the rapidly evolving landscape of humanoid robotics, the term Degrees of Freedom (DOF) often surfaces in marketing materials. However, RobotWale treats DOF counts as engineering specifications rather than marketing metrics. A high DOF count does not guarantee dexterity, nor does a low count imply poor utility. The critical metric is not the number of joints, but the torque density, control bandwidth, and the intended application environment of each actuator.

This article analyzes the current state of shipping hardware, prioritizing verified specifications from manufacturers over conceptual announcements. We examine the mechanical architecture of legs for locomotion, arms for manipulation, and hands for object interaction, with specific attention to what is actually shipping versus what remains in R&D.

Locomotion Systems: Leg DOFs Analysis

Legs are the foundation of humanoid mobility. The primary function is maintaining balance and generating forward momentum. Most shipping humanoids utilize 6 to 13 degrees of freedom per unit for the lower body.

Leg Configuration Comparisons

• Unitree H1: Utilizes a 13-DOF leg system. This includes 3 DOFs at the hip (pitch, roll, yaw), 1 DOF at the knee (pitch), and 2 DOFs at the ankle (pitch and roll). The design prioritizes robustness and torque density over anthropomorphic complexity.
• Tesla Optimus Gen 2: Early iterations showed 6 DOFs per leg (3 at the hip, 2 at the knee, 1 at the ankle). The Gen 2 update retained a similar count but introduced significant changes in actuator placement to improve efficiency. The focus is on the 'inverted pendulum' control model rather than biological mimicry.
• Figure 01: Features 12 DOFs for the full body, with a significant portion dedicated to the lower body. The legs utilize high-torque actuators designed for dynamic balance rather than static standing.
• Apptronik Apollo: A 28-DOF full body robot with 6 DOFs per leg. This configuration emphasizes stability for warehouse logistics environments.

Engineering Note: While some prototypes claim higher leg DOFs, the market standard for shipping hardware remains around 6 to 7 DOFs per leg. Adding more degrees of freedom increases the computational load for inverse kinematics without necessarily improving walking speed or payload capacity.

Locomotion Performance Metrics

Leg DOFs must be weighed against payload capacity. The Tesla Optimus is designed for internal factory logistics, carrying up to 20 kg. The Unitree H1 can carry 75 kg but focuses on agility. The Figure 01 targets warehouse logistics with a payload of 20 kg. In all cases, leg actuation is limited by thermal management and battery density.

Upper Body and Manipulation: Arm DOFs

Arm DOFs are critical for reaching and positioning payloads. Unlike legs, which are often symmetrical, arms can be asymmetrical depending on the task.

Arm Architecture Breakdown

• Tesla Optimus Gen 2: 7 DOFs per arm. This includes 3 DOFs at the shoulder (pitch, roll, yaw), 1 DOF at the elbow (pitch), and 2 DOFs at the wrist (pitch and yaw). The 7-DOF configuration allows for redundancy, meaning the hand can reach a target in multiple ways, useful for avoiding obstacles.
• Figure 01: 6 DOFs per arm. Similar to industrial robotic arms (like the UR5), but with a humanoid footprint. The lack of a wrist yaw reduces dexterity compared to the Optimus.
• Unitree H1: 6 DOFs per arm. Focuses on reach and strength rather than fine manipulation.
• Apptronik Apollo: 6 DOFs per arm. Optimized for lifting and palletizing tasks.

The Redundancy Advantage

The Tesla Optimus 7-DOF arm is a distinct advantage in cluttered environments. While a 6-DOF arm can position the end-effector at any coordinate in 3D space, it cannot easily orient the end-effector without moving the base joints. The 7th joint (usually the wrist yaw or elbow roll) allows for 'elbow up' or 'elbow down' postures to clear obstacles. However, this redundancy requires more complex control software, which is a bottleneck for current deployment.

The Hand Problem: Gripper vs. Anthropomorphic DOFs

The hand is the most contentious area in humanoid robotics. The industry is split between anthropomorphic fingers and hard-robotic grippers.

Tesla Optimus Hand

Early Optimus prototypes used a single-DOF gripper. The Gen 2 version introduced an 11-DOF hand. This includes 2 DOFs per finger (thumb and index) and 2 DOFs for the wrist. The focus is on 'grasp' rather than 'fine manipulation'. The fingers are capable of adjusting to object shapes, but they lack the independent control of human fingers.

Figure 01 Hands

Figure AI has moved towards a 3-finger gripper with 5 DOFs. This prioritizes speed and strength over dexterity. The gripper is designed for box handling and palletizing, not delicate assembly.

Unitree H1 Hands

Unitree offers a 3-finger gripper with 3 DOFs. This is a standard industrial configuration. It lacks the torque for fine manipulation but excels at heavy lifting.

Market Reality: Shipping Hardware

There is a significant gap between 'demo' hands and 'shipping' hands. Many announcements feature hands with 15+ DOFs that are not yet in mass production. Currently, the shipping standard is a hybrid approach: a 3-DOF or 5-DOF gripper for 90% of tasks, with a specific high-torque finger for a handful of exceptions.

India Availability and Pricing Estimates

The availability of humanoid robots in India is currently restricted to enterprise pilots. There is no retail channel for consumers. Import duties on robotics hardware in India are high, often ranging between 10% to 20% on the CIF (Cost, Insurance, Freight) value, plus GST of 18%.

Estimated Landing Costs

While exact prices are proprietary, market estimates suggest the following landed costs for enterprise pilots in India:

• Tesla Optimus: Estimated unit cost $25,000 to $50,000 USD. Landed cost in India could reach INR 25 Lakhs to INR 45 Lakhs depending on the pilot agreement and import classification.
• Figure 01: Estimated unit cost $200,000 USD. Landed cost in India could exceed INR 1.8 Crores. This is currently limited to large-scale manufacturing partners.
• Unitree H1: Estimated unit cost $200,000 USD. Similar pricing to Figure AI. Unitree is more accessible for R&D centers.
• Apptronik Apollo: Estimated unit cost $150,000 USD. Targeted at logistics partners.

Note: These are landed cost estimates including import duties and GST. Final pricing requires direct negotiation with manufacturers.

Regulatory Landscape in India

The Indian government classifies these robots under the 'Capital Goods' category. Import licensing may be required depending on the end-use. There are no specific regulations for humanoid safety in India yet, unlike the EU's AI Act or US OSHA guidelines. Manufacturers must comply with local electrical safety standards (BIS) for the power systems.

Conclusion: DOF Count is Not the Only Metric

As the industry moves from 2024 to 2025, the focus is shifting from 'how many DOFs' to 'how well do they work'. A robot with 40 DOFs that cannot lift a box is inferior to a robot with 20 DOFs that can lift 50 kg consistently. The shipping hardware available today prioritizes reliability, torque, and thermal management over extreme dexterity.

For Indian enterprises, the immediate opportunity lies in deployment pilots rather than direct procurement. The cost barrier remains high, and the supply chain for spare parts is not yet established locally. Manufacturers must prove long-term uptime before mass adoption can occur.

References

• Tesla AI Day 2023 Presentation - tesla.com/ai
• Figure AI Official Website - figure.ai
• Unitree Robotics Official Website - unitree.com
• Apptronik Official Website - apptronik.com
• Cabinet Secretariat, India - Import Policy for Capital Goods - dgft.gov.in