Israel’s Rafael integrates artificial intelligence into Spice bombs

By: Seth J. Frantzman and Kelsey D. Atherton

Rafael Advanced Defense System's Spice bombs now have a new technological breakthrough as the Israeli company enables its Spice 250 with artificial intelligence alongside automatic target recognition to be used with scene-matching technology.

The Spice 250, which can be deployed on quad racks under the wings of warplanes like the F-16, has a 75-kilogram warhead and a maximum range of 100 kilometers with its deployable wings.

Its electro-optic scene-matching technology — which involves uploading terrain data onto the bomb and combining it with real-time electro-optic imagery — allows the weapon to work in GPS-denied environments. And the bomb can use this autonomous capability to navigate and correct its location, according to Gideon Weiss, Rafael's deputy general manager of marketing and business development at the company's air and C4I division.

With its AI and “deep learning” technologies, the weapon has the ability to identity moving ground targets and distinguish them from other objects and terrain. This is based on 3D models uploaded to the bomb as well as algorithms. As the weapon identifies and homes in on its target, such as a convoy of vehicles, it separates the convoy of interest from other vehicles it has “learned” to ignore.

“The deep-learning algorithm is indifferent to the actual data fed to it for modeling targets of interest and embedding their pertaining characteristics into the system," Weiss said. "However, the more the data used for modeling is representative of the target of interest, the more robust the recognition probability will be in real life.”

Rafael has completed the development and testing phase of the Spice 250, including flight tests, which have “proven the robustness of the ATA and ATR, so it is mature for delivery,” Weiss said, using acronyms for automatic target acquisition and recognition.

Asked if the ATR algorithm will select a secondary target if the computer cannot find the initial human-selected target, Weiss said: “This goes into the area of user-defined policies and rules of engagement, and it is up to the users to decide on how to apply the weapon, when and where to use it, and how to define target recognition probabilities and its eventuality.”

Automatically selecting a secondary target may eventually become part of the upgrade profile for the munition, if customers express significant interest in the feature.

With a two-way data link and a video-streaming capability, the bomb can be aborted or told to re-target up until a “few second before the weapon hits its target,” Weiss explained. That two-way data-link, enabled by the weapon's mounting on a Smart Quad Rack, or SQR, will enable future deep learning to be based on data extracted from earlier launches. Data recorded will include either live-streaming video or a burst of still images of the entire homing phase up until impact.

“These are automatically and simultaneously recorded on the SQR — enabling two functions: (a) real-time and post-mission BDI (Bomb Damage Indication); (b) post-mission target data extraction for intel updates, etc.," Weiss said. "The ATR capability, including its deep learning updates, must be more agile than the enemy's ability to conceal and/or change its battlefield footprint, tactics, appearance or anything else which might impede the ATR from accurately recognizing and destroying targets.”

The Spice family of weapons is operational with the Israeli Air Force and international customers.

https://www.defensenews.com/artificial-intelligence/2019/06/17/israels-rafael-integrates-artificial-intelligence-into-spice-bombs/