The Challenges in Building Humanoid Joints Part 1: Legs & Balance
Humanoid robots, designed to emulate the human form, encounter unique challenges at each joint. Mirroring the complexity of the human body is a difficult task that we will still struggle to fully achieve. Its speed, balance, dexterity, and especially fluidity, make it particularly challenging to imitate. The inherent challenge lies not just in emulating these attributes but in the fact that human movement has been internalized to a point where analyzing it becomes daunting. For instance, an athlete might struggle to explain the mechanics of performing a backflip.
Engineers tasked with replicating these characteristics face a continually evolving set of challenges. But what exactly are these challenges? To simplify, we’ve categorized the joints by limb: upper limbs (arms), lower limbs (legs), and the neck. Each presents its distinct demands and obstacles. Let’s explore the various solutions that address these challenges.
Today in this first article of a series of three, we will talk about the challenges when it comes to building a humanoid legs robot. By focusing on its most obvious challenge: maintaining balance.
1. Humanoid Legs – the challenge of the balance
a. Bipedal Movement: a constant state of controlled falling
In the dynamic world of robotics, especially within the domain of humanoid robots, movement such as walking or running presents a profound engineering challenge that starkly contrasts with the notion of static stability. The issue of balance maintenance has been a persistent concern, as demonstrated during events like the DARPA Robotics Challenge (DRC) in 2015. Videos of the competition depicted robots frequently falling and struggling to recover their footing. When stationary, a robot may achieve a semblance of balance relatively easily.
However, the true test of its design and control systems comes to light during motion. Bipod humanoid robots are inherently unstable when standing. But that can be engineered with relative straightforwardness. Dynamic stability—required for walking or running—is a far more complex feat. This problem is not merely technical but also carries significant safety risks, especially considering the potential danger posed by a heavy humanoid robot falling on a human, such as a child.
Unlike a tripod, which remains stable without effort, a bipedal humanoid robot is in a constant state of controlled falling. Every step taken by a humanoid robot is essentially a catch of its fall, a delicate dance between imbalance and control. The robot must continuously predict, adjust, and react to a multitude of factors to simulate the relatively smooth motion observed in human walking.
In essence, while three legs would theoretically provide a humanoid robot with inherent stability, the choice to utilize two legs aims to emulate human locomotion and appearance closely. This decision significantly complicates the engineering challenge, making the quest for bipedal balance a central, ongoing puzzle in the development of humanoid robots. Many humanoid robots can now walk at a human-like speed on a more-or-less steady floor. However, the general goal is to have humanoid robots doing missions dangerous for humans (such as rescue missions). For this, they should maintain balance on uneven floors and be capable of jumping while carrying heavy weights.
b. Different joints in a Robotic humanoid leg:
Hips (A):
Actuated by powerful electric motors or hydraulic actuators, hip joints enable multi-directional movement crucial for bipedal locomotion. These joints are engineered for robustness, considering the robot’s weight distribution and balance
In humanoid robotics, the hip joint is one of the most crucial areas for enabling a wide range of motion and maintaining balance. Typically, the hip joints in humanoid robots are composed of more than one actuator. This design allows the robot to achieve movements in multiple planes, closely mimicking the human hip’s complex functions.
Human hips can move in three primary directions: forward and backward (sagittal plane), side to side (frontal plane), and rotationally (transverse plane). To replicate this range of motion, humanoid robots usually have at least three degrees of freedom (DOF) at each hip:
- Flexion/Extension: Moving the leg forward and backwards.
- Abduction/Adduction: Moving the leg away from or toward the midline of the body.
- Internal/External Rotation: Rotating the leg around its axis.
Achieving such movements typically requires multiple actuators. For example, a common setup includes:
- One actuator for flexion and extension,
- A second actuator for abduction and adduction,
- And possibly a third for internal and external rotation.
Knee (B):
Incorporating compact and durable actuators, knee joints facilitate precise flexion and extension motions, supporting dynamic weight-bearing tasks while ensuring stability. Some robots may incorporate a rolling contact mechanism to achieve a large range of motion and enhanced transportability. Here are two examples of different kinds of knee joints:
Rotary Knee Joints mimic human knee movement with a rotational actuator, ideal for replicating human walking and complex movements due to their compact design. It integrates robots incorporated between thigh and shin components.
Linear Knee Joints, though less common, use a linear actuator to extend and retract, thereby bending and straightening the knee. The mechanism converts linear motion into the angular motion required for the knee to bend. Linear actuators control knee bending with high precision and are suited for tasks requiring exact movements or operation on challenging terrains.
Note: In usual practice rotary joints are more used than linear knee joints. However, in some cases, a linear joint can be added to the calf of the robot. This lean calf actuator then helps to strengthen the rotary actuator in the ankle.
Ankle (C):
Compact servo motors or pneumatic actuators drive ankle joints, allowing for precise control over dorsiflexion, plantarflexion, and lateral movements. The design prioritizes lightweight components to minimize inertia and enhance agility. Ankle actuators play a critical role in managing the robot’s interaction with the ground, adjusting the tilt and position of the feet to keep the centre of pressure (COP) within the base of support (BOS).
Typically, an ankle joint in a humanoid robot would have at least two actuators. These are generally used to allow the robot’s foot to perform two primary movements:
- Dorsiflexion and Plantarflexion: This movement is similar to a human pointing their foot up or down. It’s crucial for walking and managing steps or uneven terrain.
- Inversion and Eversion: This movement allows the foot to tilt inward and outward, aiding in balance and adapting to various surfaces.
Some advanced humanoid robots may include additional actuators for more refined control or additional degrees of freedom, but two are common for basic models focusing on typical walking and standing tasks.
Toe Joint (D):
Designing toe joints in humanoid robots introduces a nuanced layer of complexity essential for achieving realistic and efficient bipedal movement. In the toe-off phase of walking or running, the toe joints provide the necessary push, aiding in the forward movement of the robot. This phase requires precise actuator control to emulate the force and flexibility of human toes. We can distinguish 2 kinds of toe joints: active or passive.
The distinction between active and passive toe joints reflects the level of control and actuation provided to these joints. Both types play crucial roles in the dynamics of walking, running, and balance, but they operate under different principles and serve different functions within the robotic system.
Active toe joints are equipped with their actuators, allowing for independent and precise movement control. They enhance propulsion, balance, and terrain adaptation but add complexity, weight, and energy demands to the robot.
Passive toe joints lack actuators and respond to external forces with built-in flexibility or spring mechanisms. They offer simplicity, reduce weight, and save energy but provide limited functionality and depend heavily on the robot’s overall design for effectiveness.
Choosing between active and passive toe joints depends on the robot’s intended use, with active joints suited for complex tasks and environments, and passive joints favored for simplicity and efficiency.
2. Strategies to maintain balance within robotic legs
While the broader spectrum of balance strategies encompasses everything from whole-body control to adaptive learning mechanisms. Our focus narrows to those strategies that specifically harness the strength and capabilities of actuators in the leg areas: the hips and ankles.
The hip strategy is characterized by a large rotation of the hip joint which repositions the Centre of Mass.
- Center of Mass (CoM): CoM refers to the point in a body or system of bodies where mass is evenly distributed in all directions. It’s a geometric property that doesn’t change with the body’s orientation in space. Whereas the Center of Gravity (CoG) is the point in an object or system where the gravitational force can be considered to act. In a uniform gravitational field, CoG coincides with the CoM.
The ankle strategy focuses on stabilizing the robot by applying torque at the ankle joint while keeping the hip joint locked.
These two strategies represent different approaches to maintaining balance in humanoid robots. The hips strategy relies on broader, more significant movements to control the robot’s center of mass (COM) and adjust its posture for balance. The ankle strategy focuses on smaller, finer adjustments to maintain stability, especially in relation to the robot’s interaction with the ground. Integrating Archimedes Drives can enhance the effectiveness of both methods. You can read on this topic here.
3. How can the Archimedes Drive help to achieve better humanoid balance?
In the hips strategy, Archimedes Drives can be implemented to provide additional support during hip joint rotation by providing for powerful movements, precise control with minimal backlash for accurate balance adjustments, and improved energy efficiency. By counterbalancing the weight shifts associated with hip movement, Archimedes Drives helps stabilize the robot’s Center of Mass (CoM), thereby improving balance maintenance.
Similarly, in the ankle strategy, Archimedes Drives can complement ankle torque application by providing precise, backlash-free control for fine-tuned balance adjustments, high torque density for strong and responsive movements, and energy efficiency. Crucial for the continuous, subtle corrections needed to maintain stability in humanoid robots. This ensures that the robot remains stable and upright, even when subjected to external disturbances.
