High Res TIFF (18.2 MB)

3D Stacked Neuro Processor Device:
3-Dimensional Artificial Neural Network - 3DANN

For over 30 years, technologists have been working to create machines that can see in three dimensions, and make sense of what they observe. The ability to recognize items of interest against a highly cluttered background - a common challenge in planetary landscapes - requires a highly sophisticated system that can process what it perceives, then instantly turn it into useful information.

A spacecraft headed for Mars might, for example, use a visual system to avoid dangerous terrain at a planned landing site, and help select a safer spot nearby.

Currently, machine vision system development has barely reached a level similar to the visual processing capabilities of primitive insects. Fortunately, a new development promises to greatly enhance these abilities. Helping out is a 3D-stacked neuro-processor device, known as a 3-Dimensional Artificial Neural Network or 3DANN. It was created by the Ballistic Missile Defense Organization (BMDO) and sponsored by the VIGILANTE project. Advanced research by this organization has already demonstrated a new generation of sensors and processors for the next century's missile defense systems.

3DANN has led to unprecedented strides in machine vision potential, including dynamic orientation/scale/lighting-invariant recognition of targets against highly cluttered backgrounds. Modeled after the massively parallel neural networks in human brains, 3DANN is a low-power, analog-computing device capable of achieving human-like target recognition capability.

A primary future benefit to NASA will be the creation of the visually intelligent robots necessary for effective completion of planetary exploration missions.