Ukraine's NORDA Dynamics Develops GPS-Free Navigation and AI Drone Module

According to LB.ua reporting citing company CEO Nazar Bihun, Ukrainian unmanned systems developer NORDA Dynamics is preparing to release an updated version of its serial Underdog AI-powered auto-guidance module alongside the ninth version of its fixed-wing drone algorithms.

Automated Target Recognition Powered by Neural Networks

The primary distinction of the new AI-modified Underdog compared to its base version is automated object detection. Previously, operators had to manually locate targets using optics and designate them with a capture marker; now, a neural network performs this task by recognizing objects and highlighting them on the display provided a connection between the drone and the ground control station remains active.

Based on reporting from LB.ua, company engineers utilized real combat data to train the algorithms:

• Training datasets included Brave1 Dataroom and Avengers Labs collections.
• Proprietary data gathered during field trials for specific tasks was also integrated.
• Synthetic data accounted for only a minor fraction of the training sample.

The onboard neural network model is trained to distinguish nine separate target classes. The updated optical auto-guidance system can also detect and track moving objects at distances of up to two kilometers under favorable weather conditions, though final strike authorization still requires pilot approval.

Navigation Solutions for GPS-Denied Environments

In addition to the AI targeting module, the company introduced a navigation addon designed for flights without GPS signals. The system determines routes by combining optical terrain-contour matching algorithms with map data and visual-inertial odometry. Furthermore, the system supports integration with ground-based radio beacons manufactured by Sine Engineering and Skyline.

Nazar Bihun noted that combining GPS-free navigation, neural network object recognition, and auto-guidance will serve as the foundation for fully autonomous strike missions. The company is currently finalizing the pre-release AI version of the Underdog and preparing the ninth version of fixed-wing auto-guidance algorithms for official release, targeting early 2027 for fully autonomous mission capabilities.