STRATA
Skip to study
XLVII

III.10 / INTELLIGENCE

Embodied cognition and sensorimotor learning

How does performance change when the same learner can move and sample its surroundings?

A source-led introduction1 source references1 related experiments

Start with the question.

Embodied cognition investigates how sensing, movement, and the environment participate in intelligent behavior. Brooks's robotics argument gives a practical example of organizing behavior around ongoing interaction. The relevant comparison is between architectures facing the same task and disturbances, with attention to what the body and environment make easier, harder, or unnecessary to represent explicitly.

Change the body while keeping the controller fixed.

Imagine a small obstacle-avoidance robot with sensors mounted at two different heights. Run the same controller through the same arrangement of objects and compare the paths. A changed result can come from the sensing geometry rather than an improved internal plan; the experiment makes the body part of the model.

What would you inspect?

  • Sensor placement, action latency, environmental structure, and success criteria.
  • Performance when lighting, surface friction, or obstacle position changes.

Take the next step.

Specify one sensorimotor task and compare two body or sensor configurations while keeping the controller and evaluation conditions constant.

Put part of the question to work.

These models isolate a mechanism you can vary. Their assumptions define the part of this topic they can illuminate.

Try forcing, damping, and response.

Change one parameter, compare the numerical output, and export the result. The experiment loads when you open it.

Open its full page

Watch the quantities change.

12 seconds · Original model visualization · Silent

Forcing, damping, and response

Compare a finite driven motion with the response that persists after transients decay.

Read the explanation and method

A damped oscillator starts at rest and is driven sinusoidally at its natural frequency of 1 hertz. The damping ratio is 0.12.

Cream shows normalized displacement and orange shows the normalized applied force. The transient grows toward a periodic response.

The analytical steady-state gain is about 4.17. The displayed finite trajectory comes from fourth-order Runge-Kutta integration and may retain a transient.

Method. Integrate the normalized mass–spring equation with fourth-order Runge–Kutta steps of 1/240 second. The plotted trace keeps 60 samples per second. Separately calculate the analytical steady-state gain and phase for a continuously driven linear system.

Read the supporting material.

  1. Intelligence without Representation

    Rodney A. Brooks / MIT · The author's embodied, situated robotics argument and architecture.

Follow the comparisons.

Read the neighboring questions.