Actuator Overview

The actuator system in MuJoCo provides a unified mechanism for applying forces, torques, or generalized controls to joints and tendons. Actuators are essential for driving motion, simulating motors, springs, muscles, hydraulic cylinders, and a variety of biomechanical or robotic control systems.

This section explains:

  • How actuators work conceptually

  • How they apply forces within MuJoCo’s dynamics

  • The structure and purpose of the main actuator shortcuts: motor, position, velocity, intvelocity, damper, cylinder, muscle, adhesion

No attribute-level details are included—only structural and functional concepts.


1. What an Actuator Does

In MuJoCo, an actuator is a mechanism that maps control signals (inputs) into forces or torques that act on:

  • a joint, or

  • a tendon, or

  • a slider-like constraint such as in hydraulic actuators

Actuators allow you to:

  • Drive a robot’s joints

  • Apply force through cables, tendons, or motors

  • Implement closed-loop or open-loop controllers

  • Simulate biologically inspired actuators (muscles, springs)

  • Add damping, stiffness, adhesion, or constraint-driven forces

Core Logic

  1. The simulation receives a scalar control signal (e.g., ctrl[i]).

  2. The actuator translates this signal into a scalar generalized force.

  3. This force is applied to the controlled joint/tendon in generalized coordinates.

  4. The solver integrates the result into the system dynamics.

Thus, actuators bridge control input → physical force.


2. Actuator Shortcuts Overview

MuJoCo provides a collection of high-level “shortcut” actuator types that correspond to common control behaviors. These are convenience wrappers around the more general actuator model, each configured for a specific dynamic behavior.

Below is an overview of each actuator shortcut and its structural role.


2.1 Motor

A motor actuator applies force or torque proportional directly to the control input.

  • Represents ideal torque or force motors

  • Most common actuator for robotic joints

  • Linear mapping from control → force

Structurally: 👉 The controller directly determines the joint or tendon effort.


2.2 Position

The position actuator acts like a servo that drives a joint toward a target angle or displacement.

Characteristics:

  • Accepts a desired position as control

  • Internally generates a force that pulls toward that target

  • Implements proportional (and often soft) position control

Structurally: 👉 A closed-loop controller hidden inside the actuator.


2.3 Velocity

The velocity actuator tries to regulate joint velocity toward a target.

  • Control input = desired velocity

  • Actuator produces force to reduce velocity error

  • Useful for speed-controlled motors or rotational drives

Structurally: 👉 A damping-like force with velocity tracking.


2.4 IntVelocity (Integrator Velocity)

The intvelocity actuator integrates the control input over time to produce a target velocity.

  • Control signal is integrated → becomes a velocity target

  • Smoothly accumulates input

  • Enables “rate-based” velocity control

Structurally: 👉 Suitable for controllers where velocity should evolve smoothly.


2.5 Damper

A damper actuator resists relative motion by applying force proportional to velocity.

  • Produces viscous damping

  • No control input needed

  • Helps stabilize joints or absorb shocks

Structurally: 👉 Passive element that adds velocity-dependent resistance.


2.6 Cylinder

A cylinder actuator simulates:

  • Pneumatic cylinders

  • Hydraulic pistons

  • Linear actuators with pressure/force based on displacement

Functional characteristics:

  • Applies force along a defined line between two bodies

  • Length of the cylinder depends on body positions

  • Control input often interpreted as pressure or force

Structurally: 👉 Force applied based on geometric extension.


2.7 Muscle

The muscle actuator implements a biomechanical muscle model.

  • Computes force based on activation dynamics

  • Includes force–length and force–velocity relationships

  • Control input often represents muscle activation level

Structurally: 👉 Physiology-inspired actuator for tendons and musculoskeletal systems.


2.8 Adhesion

The adhesion actuator models adhesion or sticking forces between surfaces.

  • Applies attractive force to keep two sites together

  • Used for simulating suction, sticky pads, climbing robots

  • Can act like a controllable “sticky” constraint

Structurally: 👉 Applies a force that resists separation between two points.


3. Actuator System Summary

MuJoCo’s actuator system provides a versatile way to convert control signals into physical forces applied to joints or tendons. Actuators support a diverse range of use cases:

  • Robotic motors and servos

  • Damping and stabilization

  • Hydraulic and pneumatic cylinders

  • Biomechanical muscle models

  • Adhesive or suction-like attachment forces

Shortcut types (motor, position, velocity, intvelocity, damper, cylinder, muscle, adhesion) offer convenient configurations for commonly needed actuator behaviors.

Together, they allow MuJoCo to simulate a wide variety of robotic actuators, biological muscles, compliant elements, and force-based mechanisms in a unified dynamic framework.


Last updated