This post is also available in:
עברית (Hebrew)
GPS has become fundamental to modern drones and guided systems, but it also creates a vulnerability. Satellite signals can be jammed or spoofed, potentially leaving an unmanned aircraft unable to determine where it is or continue accurately toward its destination.
A new autonomous drone prototype called Matador is being developed around a different approach: navigate by looking at the terrain instead.
Developed by Morocco’s Ballistic Defense System, the drone is envisioned as a long-range, one-way attack drone capable of operating without depending on GPS. All of its processing takes place onboard using a computer approximately the size of a human hand.
Rather than relying on one AI system for the entire flight, it divides the mission between multiple specialized AI agents.
One handles navigation, another searches for the target and a third manages the final approach. The architecture is designed to activate different capabilities according to the stage of the mission.
According to NextGenDefense, alongside them is what the developer describes as a “super agent” based on Kalman filtering. A Kalman filter is a mathematical method commonly used to estimate a system’s state by combining imperfect measurements. In the drone, it weighs available information to help determine the aircraft’s course of action.
The navigation system relies heavily on vision.
Cameras observe the terrain beneath and around the drone, and the onboard system compares those observations with a freely available global elevation map. By matching what the cameras see with known terrain features, the aircraft can estimate its location without requiring continuous satellite-navigation signals.
That capability is particularly relevant to electronic warfare. GPS jamming can deny navigation signals altogether, while spoofing attempts to feed a receiver false positioning information. A drone capable of navigating from visual terrain information has an alternative reference when satellite data becomes unreliable.
The prototype was developed in less than 18 months, with a reported budget below $1.15 million. According to the developer, the aircraft was designed, programmed, assembled and tested domestically without outsourcing.
The company says initial navigation tests compared favorably with published research from leading universities, although detailed performance figures have not been provided.
The drone will next move into live demonstrations under real-world conditions, with the results expected to guide subsequent versions.
The concept reflects a broader change in autonomous warfare. Resilience no longer depends only on making communications or GPS receivers harder to jam. Another option is to reduce how much the machine needs those external signals in the first place.
For the drone, that means giving the drone its own eyes, onboard processing and several specialized AI systems, and asking them to work out where it is and where it needs to go.
