Impedance Control

Impedance control does not directly command force or position. Instead, it regulates the dynamic relationship between the robot's end-effector position (or velocity) error and the resulting contact force. This is mathematically defined by a target mechanical impedance, typically modeled as a mass-spring-damper system: F = M * (ẍ_d - ẍ) + B * (ẋ_d - ẋ) + K * (x_d - x), where M, B, and K are the desired inertia, damping, and stiffness matrices. The controller makes the robot behave as if it had these virtual mechanical properties.

A defining feature is inherent force reactivity. When the robot encounters an unexpected object or force, the controller allows deviation from the planned trajectory according to the defined impedance. This creates passive compliance.

• High Stiffness (K): The robot resists deflection, behaving rigidly to track position precisely in free space.
• Low Stiffness (K): The robot yields easily to contact forces, ensuring safe interaction and absorbing impacts.
• Damping (B): Critically dampens the response to prevent oscillations upon contact or release. This is crucial for tasks like assembly, where a rigid peg-in-hole search would cause jamming, while a compliant search allows chamfers to guide alignment.

Impedance control sits between two classical paradigms:

• Position Control: Commands a precise trajectory. On contact, it fights to maintain position, generating high, potentially damaging forces. It has infinite impedance.
• Force Control: Commands a specific force. In free space, it can "push" into nothing, causing unstable runaway motion. It has zero impedance.

Impedance control unifies these by specifying a finite, tunable impedance. It can prioritize position tracking in free space (like position control) while gracefully accommodating contact (like force control), making it ideal for unstructured environments where contact is uncertain.

The behavior is tuned via symmetric, positive-definite matrices:

• Stiffness Matrix (K): Defines the restorative force per unit of positional error. A diagonal K with high values creates a stiff, precise Cartesian spring along each axis. Off-diagonal terms create cross-coupling, e.g., a lateral force causing a rotational deflection, useful for complex tool interactions.
• Damping Matrix (B): Defines the dissipative force per unit of velocity error. Proper damping is critical for contact stability. It is often chosen to achieve critical damping for the virtual mass-spring system, preventing bouncing when making or breaking contact. These matrices can be anisotropic (different values per axis) and configuration-dependent, allowing a robot to be stiff in one direction (e.g., drilling) and compliant in another (e.g., surface following).

Practical implementation requires a torque-controlled robot. The general process is:

1. Measure the actual end-effector position x and external force F_ext (via a force/torque sensor or joint torque sensing).
2. Calculate the desired force F_des using the impedance law and the error from the reference trajectory x_d.
3. Use inverse dynamics to compute the joint torques τ required to generate F_des at the end-effector, accounting for the robot's own dynamics (gravity, Coriolis forces).
4. Command τ to the joint-level torque controllers. This closed-loop force modulation is computationally intensive and requires an accurate dynamic model, but it enables the emergent compliant behavior.

In legged robots, impedance control is fundamental for managing ground interaction.

• Leg Compliance: Each leg acts as a virtual spring-damper, absorbing impacts at touchdown and storing/releasing energy for efficient gaits, inspired by the Spring-Loaded Inverted Pendulum (SLIP) model.
• Adaptation to Uneven Terrain: Upon foot contact with an unexpected height, the leg compresses according to its impedance instead of exerting a large force to reach the pre-planned position, providing intrinsic terrain adaptation.
• Whole-Body Impedance: The robot's torso or whole body can be assigned an impedance, allowing it to "lean into" pushes or absorb disturbances through coordinated, compliant motion of all joints, aiding push recovery. It is often used in conjunction with higher-level planners that specify the desired foot placement and body trajectory.

For legged robots, impedance control is applied at each leg joint or foot to manage Ground Reaction Forces (GRF) and absorb impacts. Key uses include:

• Impact Absorption: A low stiffness at touchdown reduces shock on the mechanical structure.
• Terrain Conformity: On uneven ground, the leg compresses like a spring, allowing the foot to maintain full contact and stability without precise terrain mapping.
• Energy Efficiency: In combination with Series Elastic Actuators (SEA), it enables passive energy storage and return in gaits like running, mimicking the Spring-Loaded Inverted Pendulum (SLIP) model.

These material removal tasks require maintaining a consistent contact force against a surface with unknown geometry. A force-controlled impedance regulator adjusts the tool's position in real-time based on measured force error. If the force is too high, the robot retracts; if too low, it presses in. This ensures uniform material removal without gouging the workpiece, compensating for part tolerances and tool wear far more effectively than pure position control.

Impedance control enables bimanual manipulation of flexible or delicate objects (e.g., cloth, a hose) by coordinating the compliance of two arms. In haptic teleoperation, it is used in two key ways:

• Master Side: Provides force feedback to the human operator based on slave robot/environment interactions.
• Slave Side: Implements compliance on the remote robot to ensure safe interaction with its environment. The controller's ability to shape the force-position dynamic is essential for transmitting a realistic sense of touch and remote dexterity.

Admittance control is a complementary, force-reactive strategy to impedance control. Instead of regulating the force resulting from a position error (as in impedance control), admittance control regulates the motion resulting from a measured force. The controller accepts an external force as input and outputs a desired velocity or position command.

• Key Distinction: Impedance control is typically implemented in torque-controlled actuators, while admittance control is often used with position-controlled hardware.
• Application: Used in industrial cobots where a human can physically guide the robot's arm; the measured interaction force is converted into a smooth motion command.

Series Elastic Actuation (SEA) is a hardware design paradigm that physically embodies the principles of impedance control. A compliant elastic element (like a spring) is placed in series between the motor and the robot's output link.

• Mechanism: The spring intentionally decouples the motor inertia from the output, providing inherent mechanical compliance and excellent force sensing through spring deflection.
• Benefits: Enables high-fidelity force control, absorbs impact shocks, and can store/release energy for efficient dynamic motions like running.
• Example: Boston Dynamics' Atlas robot uses series elastic actuators in its legs for dynamic, force-controlled locomotion.

Whole-Body Control (WBC) is a hierarchical optimization framework that coordinates all of a robot's joints to execute multiple tasks simultaneously while respecting physical constraints. Impedance control is often used to define the dynamic behavior for individual contact or manipulation tasks within the WBC hierarchy.

• Core Function: Solves a quadratic program (QP) at a high rate (e.g., 1 kHz) to compute optimal joint torques or accelerations.
• Task Examples: Maintain balance (centroidal dynamics), track foot swing trajectories (impedance), avoid joint limits, and minimize energy consumption.
• Integration: The WBC optimizer can generate desired contact forces, which are then realized by low-level joint torque controllers implementing the specified impedance.

Ground Reaction Force (GRF) is the total force vector (normal + friction) exerted by the ground on a robot's foot. The Center of Pressure (CoP) is the point on the contact surface where this resultant force is applied.

• Fundamental Relationship: Impedance control directly regulates the GRF at each foot. By modulating stiffness and damping, the controller shapes how the foot interacts with the ground.
• Stability: The location of the CoP relative to the support polygon (the convex hull of all contact points) is critical for balance. Impedance control helps keep the CoP within stable bounds by managing force distribution.
• Measurement: Typically estimated via force/torque sensors in the feet or ankles.

Model Predictive Control (MPC) is an advanced, optimization-based control method that uses an internal dynamic model to predict future system behavior over a finite time horizon. It is frequently used in tandem with impedance control for locomotion.

• Typical Architecture: A high-level MPC plans optimal body trajectories and future Ground Reaction Forces several steps ahead. A low-level impedance controller then executes these plans, translating the desired forces into joint torques and providing compliant, reactive behavior at the contact level.
• Synergy: MPC provides anticipatory, optimal planning, while impedance control provides robust, reactive disturbance absorption at the moment of execution.

A Reduced-Order Model (ROM) is a simplified mathematical representation that captures the essential dynamics of a complex legged robot for planning and control. Impedance control parameters are often derived from or tuned based on these models.

• Common Examples:
  • Linear Inverted Pendulum Model (LIPM): For walking; assumes constant center of mass height.
  • Spring-Loaded Inverted Pendulum (SLIP): For running/hopping; models the leg as a massless spring.
• Purpose: These models make trajectory optimization (e.g., for MPC) tractable. The desired compliant behavior from impedance control is designed to make the full, complex robot approximate the stable dynamics of its simpler ROM.