Impedance Control

Impedance control does not directly command forces or positions. Instead, it defines a desired dynamic relationship between the robot's end-effector motion and the contact forces it experiences. This relationship is mathematically modeled as a second-order mass-spring-damper system. The controller's parameters—inertia (M), damping (B), and stiffness (K)—are programmed to make the robot behave as if it were a virtual physical object with those properties. For example, a high-stiffness setting makes the robot behave like a rigid wall, while a low-stiffness setting makes it compliant like a spring.

The core equation governing impedance control is derived from Newton's second law for the virtual system: F = Mẍ_d + B(ẋ_d - ẋ) + K(x_d - x). Here, F is the measured interaction force, x is the actual end-effector position, and x_d is the desired position. The controller's goal is to make the robot's actual behavior match this equation. Unlike admittance control (which computes motion from force), impedance control typically computes the joint torque required to achieve this force-motion relationship, making it well-suited for direct-drive actuators.

A primary advantage of impedance control is inherent mechanical compliance. By programming a low stiffness (K), the robot will naturally yield upon unexpected contact, making it safer for human-robot collaboration and more robust in unstructured environments. This contrasts with traditional position control, which fights against external forces to maintain a rigid pose, potentially causing damage. This compliance is crucial for tasks like assembly, polishing, or physical guidance, where the robot must adapt to contact geometry and uncertainties.

A critical challenge in impedance control is maintaining stable interaction with stiff environments. When a low-impedance robot contacts a very rigid surface (like a wall), the rapid change in dynamics can cause contact instability, leading to oscillations or bouncing. Stability analysis involves considering the robot's intrinsic dynamics, sensor delays, and the environment's stiffness. Techniques to ensure stability include:

• Passivity-based control to guarantee energy dissipation.
• Adjusting damping (B) based on estimated environment stiffness.
• Using force sensors or torque sensing at the joints for accurate feedback.

Impedance control is implemented in two primary architectures:

• Torque-Based Impedance Control: The desired impedance law is solved to compute a required joint torque, which is sent directly to the robot's torque-controlled actuators. This method offers high-fidelity impedance rendering but requires precise joint-level torque sensing or estimation.
• Position-Based Impedance Control (or Inner Position Loop): The impedance law is solved to compute a modified motion trajectory. This trajectory is then sent to a fast, high-gain inner position controller. This is more common on industrial robots without direct torque control but can be less accurate in rendering the desired dynamics due to the bandwidth limitations of the inner loop.

Impedance control is fundamental to advanced dexterous manipulation. Its ability to manage contact forces makes it essential for:

• In-hand manipulation: Modulating grip forces and finger compliance to roll or slide an object within a hand.
• Tactile servoing: Using tactile sensor feedback to adjust impedance parameters for following contours or maintaining stable contact.
• Non-prehensile manipulation: Controlling the compliance of a palm or finger to push or pivot an object.
• Physical human-robot interaction (pHRI): Allowing a human to physically guide the robot by "leading" its compliant end-effector.

Admittance control is the dual strategy to impedance control. Instead of regulating force in response to motion (impedance), it regulates motion in response to measured force. The controller takes an external force measurement as input and outputs a desired motion or velocity command.

• Key Difference: Impedance = Motion → Force. Admittance = Force → Motion.
• Typical Use Case: Often implemented in hardware where a force/torque sensor is available at the wrist, as it directly uses force feedback.
• System Behavior: Makes the robot's end-effector behave as if it is connected to the environment through a programmable spring-damper system, but the causality is reversed compared to impedance control.

A Series Elastic Actuator is a hardware embodiment of the principles behind impedance control. It places a compliant element (like a spring) in series between the motor and the output link.

• Mechanical Impedance: The physical spring provides inherent, passive compliance, acting as a natural force sensor through Hooke's Law (force is proportional to spring deflection).
• Benefits: Enables high-fidelity force control, excellent shock absorption, and increased safety during human-robot interaction.
• Relationship to Control: While impedance control is a software strategy, an SEA provides a complementary hardware foundation that simplifies achieving compliant, force-sensitive behavior.

Force closure is a fundamental condition for a secure grasp in robotics. A grasp achieves force closure if the set of contact forces the robot can apply can generate any resultant wrench (combined force and torque) on the object, thereby resisting arbitrary external disturbances.

• Mathematical Guarantee: It ensures the object can be held immobile even if pushed or twisted.
• Connection to Impedance: Once a force-closure grasp is established, impedance control at the wrist or finger joints can be used to manage the interaction forces between the grasped object and the environment (e.g., inserting a peg, turning a crank).
• Analysis Tool: Often analyzed using the Grasp Wrench Space, which visualizes all possible wrenches applicable through the contacts.

Model Predictive Control is an advanced, optimization-based control method highly relevant for dynamic manipulation. It uses an internal dynamic model of the robot and its environment to predict future states over a horizon and solves for an optimal sequence of control inputs.

• Integration with Impedance: MPC can be used to generate reference trajectories for an impedance controller. More advanced formulations can directly optimize for desired impedance parameters (stiffness, damping) as part of the cost function to achieve tasks like catching or striking.
• Handles Constraints: Excellently manages physical limits (joint torque, velocity) and contact constraints, which are central to dexterous manipulation.

The Jacobian matrix is the critical mathematical tool that bridges joint space and Cartesian (task) space, making it essential for implementing impedance control.

• Function: For a robot manipulator, the Jacobian (J) linearly relates joint velocities (q̇) to the end-effector's Cartesian linear and angular velocity (v): v = J q̇.
• Role in Impedance Control: It is used to transform the desired Cartesian impedance (forces in task space) into the required joint torques. The fundamental control law often involves the Jacobian transpose: τ = Jᵀ F_desired, where F_desired is the force computed by the impedance controller.
• Dynamic Version: The Jacobian derivative is also needed to account for Coriolis and centrifugal forces in dynamic implementations.

Gravity compensation is a foundational layer in most torque-controlled robots, including those using impedance control. It is the process of calculating and applying the exact joint torques needed to counteract the weight of the robot's own links.

• Purpose: It allows the controller to focus on managing dynamic interactions with the environment without fighting gravity. The robot arm can float effortlessly as if in zero gravity when no other commands are given.
• Prerequisite for Impedance: Effective impedance control typically requires accurate gravity compensation as a baseline. The impedance control law then adds torques on top of this to achieve the desired mass-spring-damper behavior relative to the gravity-neutralized state.
• Implementation: Relies on a good dynamic model of the robot's mass properties.