Quasi-Direct-Drive Motors: The Backdrivable Joint Revolution in Humanoid Robotics

What Is Quasi-Direct-Drive?

Quasi-Direct-Drive (QDD) motors represent a significant shift in the actuation philosophy for advanced robotics, particularly within the humanoid sector. Unlike traditional actuation chains that rely on high-ratio gearboxes (often 10:1 to 100:1) to amplify motor torque, QDD systems utilize direct-drive or near-direct-drive topologies with low reduction ratios (typically 1:1 to 1:5). This architectural choice prioritizes backdrivability—the ability for external forces to move the joint manually against the motor's electromagnetic resistance—over raw torque multiplication.

In a standard harmonic drive or cycloidal reducer setup, the gearbox acts as a mechanical barrier. If a robot falls, the high gear ratio prevents the motor from being turned by the payload, making the system stiff and potentially dangerous in human environments. QDD motors, often utilizing high-torque inner-rotor or outer-rotor designs with custom magnets, eliminate this barrier. The result is a joint that feels compliant, allowing for safe physical interaction and energy-efficient movement through passive dynamics.

Technical Architecture and Control

The core of QDD technology lies in the motor design. To achieve high torque without a gearbox, manufacturers increase the number of poles and the magnet surface area. For example, a QDD actuator might utilize a 12-pole motor with high-temperature resistant magnets to sustain torque at low speeds. This requires high-fidelity control loops. Because there is no gearbox filtering out high-frequency noise or torque ripple, the control algorithm must be exceptionally precise to prevent jitter.

Control systems for QDD motors typically rely on Field-Oriented Control (FOC) with high sampling rates, often exceeding 10 kHz. This allows for torque control modes where the motor resists or yields to force inputs. In contrast, traditional servo systems often operate in position or velocity modes, where the controller fights back against any deviation from a target position. In QDD systems, the controller modulates current to match the external force, creating a virtual damping effect.

The Safety Case for Backdrivable Joints

Backdrivability is not merely a technical specification; it is a safety requirement for humanoid robots operating in proximity to humans. When a humanoid robot interacts with a person, unpredictable forces occur. If a joint is non-backdrivable and the robot loses balance, the stored energy in the gearbox or the resistance of the motor can cause injury. A backdrivable QDD joint allows the human to physically resist the robot's motion or push the robot out of harm's way.

Furthermore, QDD systems reduce the risk of mechanical failure. Gearboxes are prone to wear, backlash, and catastrophic failure under shock loads. By removing the gearbox, the transmission path is shortened. There is no mechanical slack to absorb energy, meaning the motor must handle the shock. This places a higher burden on the motor's structural integrity and the control software's ability to handle sudden load changes.

Current Commercial Implementations

As of late 2024, the adoption of QDD technology is concentrated among companies shipping hardware for industrial or pilot use. Several key players have moved beyond conceptual renderings to deliver functional prototypes.

Figure AI and the Figure 01

Figure AI has publicly demonstrated the Figure 01 robot, which reportedly utilizes high-torque actuators in the arms and legs. While the company has not released a full spec sheet detailing the reduction ratios of every joint, their demonstration of the robot lifting heavy objects and performing complex tasks suggests a drive system capable of high torque density. The focus on dexterity implies a shift away from traditional industrial servo stacks toward custom actuator designs.

Agility Robotics and the Digit

Agility Robotics, known for the Digit bipedal robot, has pioneered actuator design for dynamic motion. The Digit robot uses proprietary actuators that are not traditional DC motors with planetary gearboxes. Their 2023 updates to the Digit platform indicated a move toward higher torque density and improved energy efficiency. The Digit's ability to carry loads up to 90kg suggests a QDD-like architecture where the leg actuators must handle high impulse loads without mechanical transmission failure.

Unitree and the H1/G1 Platforms

Unitree Robotics has gained significant traction for its H1 humanoid, which features 41 degrees of freedom. The H1 specifications indicate the use of high-torque motors in the legs with minimal gearing. The G1 model, launched as a more accessible entry, continues this trend with a focus on cost-effective QDD implementation. Unitree's factory videos show the joints moving under load with significant compliance, suggesting the presence of QDD technology in their core actuation stack.

Apptronik and the Apollo

Apptronik’s Apollo robot, focused on warehouse logistics, utilizes a full-body actuation system designed for durability. While specific reduction ratios are often treated as trade secrets, the robot's ability to manipulate cargo in tight spaces implies a compliance layer that aligns with QDD principles. The emphasis on durability suggests that the torque is managed through the motor's electromagnetic properties rather than mechanical advantage.

Trade-Offs: Cost, Thermal, and Control

Despite the advantages, QDD motors are not a universal solution. They come with significant trade-offs that manufacturers must manage.

Cost and Manufacturing Complexity

QDD motors require rare earth magnets (such as Neodymium Iron Boron) and precision windings to match the torque output of a gear-reduced motor. This increases the Bill of Materials (BOM) cost. A single high-torque QDD joint can cost significantly more than a standard servo motor with a harmonic drive. For a humanoid robot with 40+ joints, this cost differential is substantial.

Thermal Management

Without a gearbox to reduce the speed of the motor relative to the output shaft, the motor windings must handle higher currents for longer durations. This generates heat. In a humanoid robot, heat dissipation is a challenge due to the compact form factor. Manufacturers often employ liquid cooling or advanced heat sinking within the joint housing to prevent thermal throttling during continuous operation.

Control Software Complexity

As mentioned, the control loop must be robust. If the sensor noise (encoder resolution) is not high enough, the backdrivability can result in oscillations. The control stack must account for friction, gravity, and payload changes in real-time. This requires advanced software engineering, often involving model-based control strategies rather than simple PID loops.

QDD Availability in India

For Indian robotics startups and research institutions, the adoption of QDD technology presents specific challenges. Most QDD actuators are manufactured in the US or China, subjecting them to import duties and logistics costs.

Pricing Estimates (INR)

While exact pricing is rarely public, we can estimate landed costs based on component BOMs and shipping data.

• Standard Industrial Servo (with Gearbox): Approx. ₹20,000 to ₹50,000 per joint.
• QDD Actuator (High Torque): Approx. ₹1,00,000 to ₹3,00,000 per joint.
• Complete Humanoid Kit: A full set of QDD joints for a 1.5m robot could exceed ₹10,00,000 to ₹15,00,000 in landed cost.

These estimates include a 15-20% customs duty on electronics and heavy machinery imports. For Indian startups, this high barrier to entry often necessitates partnerships with overseas OEMs or the development of simplified local actuation solutions.

Research & Development Landscape

Despite the cost, Indian academic institutions are beginning to explore QDD concepts. IIT Madras and IIT Bombay robotics labs have prototyped low-torque density backdrivable joints for educational purposes. These are often custom-wound motors rather than commercial off-the-shelf units. Startups like Roboviz and others in the humanoid space are exploring QDD for compliance, but mass production remains a hurdle due to the supply chain for high-grade magnets.

Conclusion

Quasi-Direct-Drive motors are not merely a trend; they are a necessary evolution for humanoid robots intended to operate in shared spaces. The shift from torque multiplication to torque control fundamentally changes how robots interact with the physical world. While the cost and thermal challenges are significant, the safety benefits are non-negotiable for commercial deployment.

As manufacturing scales, we expect the cost of QDD motors to decrease. For now, they remain the domain of well-funded organizations and pilot deployments. For the Indian market, the focus remains on understanding the control architecture and adapting to the high capital cost of importing these actuators.

References

1. Figure AI. (2024). Figure 01 Robot Specifications and Demo. Retrieved from figure.ai
2. Agility Robotics. (2023). Digit Actuator Technology Whitepaper. Retrieved from agilityrobotics.com
3. Unitree Robotics. (2024). H1 Humanoid Robot Technical Data. Retrieved from unitree.com
4. Apptronik. (2024). Apollo Logistics Robot Overview. Retrieved from apptronik.com