Implemented

Vision and sensing

A tracking system has to recover a moving object after it leaves view. A binocular model lets us separate what the observer saw from what it predicted.

Conceptual model

A target appears from two viewpoints. Changing gaze changes where the target falls in the next observation.

  1. TargetAn object moves in the scene.
  2. Two viewsEach sensor samples a projection.
  3. Gaze updateThe next observation tests the estimate.

Reference function

How can an observer direct its gaze and correct a prediction when fresh visual evidence arrives?

An object projects to different positions in two views. Rotating the eyes changes those observations. While the object is hidden, a controller must carry its estimate forward without new pixels.

How the model works

Scripted scenes provide binocular images and controllable gaze. The workbench compares predicted and observed target locations. A separate synthetic sensor models adaptation and timed neural events.

SIML Biological Models →

What the model calculates

Target image position
pixels
Gaze orientation
angle
Elapsed simulation time
seconds

Evidence and limitations

What has been checked

Recorded synthetic scenes exercise tracking, disappearance, recall and correction when an object returns to view. Replay checks reproduce the recorded sequence of observations and decisions.

What remains uncertain

The scenes and sensors are engineered. Independent human gaze and retinal recordings would be needed to establish how faithfully the model predicts a person’s response.

Investigational

How could we test an alternative?

Cameras on a movable rig could reproduce a selected tracking function using a different sensor and controller.

A functional comparison

Show both systems a target that disappears briefly and changes direction. Compare reacquisition time and angular error across trajectories withheld during development.

Next validation milestone

Specify a shared target trajectory and sampling clock, then compare tracking error and gaze timing with independent observations.

Discuss a modeling collaboration