Admittance Control
Bridge the gap between rigid automation and safe human collaboration. Admittance control translates external forces into compliant motion, allowing AGVs to interact naturally with their environment and human operators.
Core Concepts
Force Inputs
The system measures external forces applied to the robot chassis using force-torque sensors or motor current estimations, serving as the primary control input.
Virtual Mass (Inertia)
Simulates the physical weight of the robot. A higher virtual mass makes the AGV feel heavier and harder to accelerate, smoothing out jittery inputs.
Virtual Damping
Acts as a stabilizing force (like moving through honey). It prevents oscillations and ensures the robot stops drifting once the external force is removed.
Velocity Output
Unlike impedance control (which outputs force), admittance control calculates the desired velocity trajectory based on the Mass-Spring-Damper model.
Compliance
The ability of the robot to yield to collisions or guiding hands. This is essential for collaborative environments where robots share space with people.
Stability Control
Advanced algorithms ensure that the virtual dynamics do not create positive feedback loops, maintaining robot stability even during hard contacts.
How It Works
Admittance control functions by reversing the traditional causality of mechanical interaction. Instead of the robot dictating a position and fighting to maintain it, it measures the forces applied to it and calculates how a virtual object would move under those forces.
The core mathematics rely on the Mass-Spring-Damper equation:
F_ext = M(a) + D(v) + K(x)
.
The controller solves this equation in real-time to generate velocity commands. For example, if a human pushes the AGV, the sensor detects the force, and the admittance filter calculates a velocity vector that moves the robot with the push, creating a sensation of weightlessness or power steering.
Real-World Applications
Hand-Guiding & Teaching
Operators can physically grab an AGV or mobile manipulator and guide it through a new path. The robot records these positions for future playback, making programming intuitive without writing code.
Heavy Payload Positioning
In assembly lines, workers may need to effect fine adjustments to a heavy chassis carried by an AGV. Admittance control amplifies the worker's small force inputs to move tons of weight effortlessly.
Crowded Navigation
In dynamic environments like hospitals, if a robot bumps into a person or obstacle, admittance control allows the robot to absorb the shock and stop/retreat immediately, preventing injury.
Docking & Coupling
When an AGV docks with a charging station or a conveyor, slight misalignments are inevitable. Admittance control provides the necessary mechanical "give" to ensure a secure latch without damaging connectors.
Frequently Asked Questions
What is the difference between Admittance and Impedance control?
The difference lies in causality. Impedance control measures motion (position/velocity) and outputs a force (ideal for lightweight, back-drivable robots). Admittance control measures force and outputs motion (velocity/position), making it ideal for stiff, heavy, non-back-drivable platforms like most industrial AGVs.
What hardware is required to implement admittance control?
Minimally, you need a method to estimate external forces. This is best achieved with a multi-axis Force-Torque (F/T) sensor mounted to the chassis or handle. Alternatively, some systems estimate force using motor current sensors, though this is less accurate due to friction.
Why is my robot oscillating or vibrating when touched?
This is usually due to low virtual damping or a control loop latency that is too high. Increasing the virtual damping parameter (D) or ensuring your control loop runs at a sufficiently high frequency (typically >100Hz) usually resolves instability.
Can admittance control work on sloped surfaces?
Yes, but it requires gravity compensation. The internal model must subtract the force vector generated by gravity on the slope from the sensor readings; otherwise, the robot will "feel" a force pulling it downhill and drift downwards continuously.
How does this impact the AGV's battery life?
Impact is generally negligible during idle states. However, if the robot is frequently used in "hand-guiding" mode, it may consume slightly more power than autonomous navigation because the motors are constantly reacting to dynamic, unoptimized human inputs rather than computed efficient paths.
Is admittance control safe for ISO 3691-4 compliance?
Admittance control is a software feature and must be paired with safety-rated hardware (PLd/Category 3) to meet safety standards. While it enhances safety by reducing impact forces, the admittance loop itself is rarely safety-rated; you still need safety scanners and emergency stops.
What is "Virtual Stiffness" and when should I use it?
Virtual stiffness acts like a spring trying to return the robot to a center point. You use it when you want the robot to resist displacement (e.g., holding a position while a worker tightens a bolt) but still yield if the force becomes dangerous.
How do I tune the Mass, Damping, and Stiffness parameters?
Start with high damping and mass for a sluggish, safe feel. Gradually reduce mass to make it lighter to push, and reduce damping to make it glide further. Ensure damping is critically damped relative to stiffness to avoid overshoot.
Does this work with Mecanum or Omni-directional wheels?
Absolutely. Admittance control is particularly impressive on omni-directional platforms because it allows operators to push the robot sideways or rotate it in place effortlessly, fully utilizing the holonomic capabilities.
What happens if the force sensor fails?
If a sensor drifts or disconnects, it might report a "ghost force," causing the robot to drift. Robust implementations include a "deadman switch" (enable button) on the handle, ensuring the robot ignores sensor inputs unless the operator is intentionally engaging it.
Can I combine this with autonomous navigation?
Yes. This is called "shared control." The robot can follow a path autonomously, but an operator can apply force to deviate from the path (e.g., to avoid an obstacle the robot didn't see), with the robot blending both inputs seamlessly.
Is it expensive to add to an existing fleet?
The primary cost is the Force/Torque sensor, which can range from $2,000 to $5,000 per robot. However, for basic applications, current-based sensing (sensorless) requires only software updates, costing significantly less to implement.