In a major leap for autonomous flight, researchers have developed an artificial intelligence framework that allows drones to navigate accurately without relying on GPS or cameras. Instead, the system uses onboard sensors such as LiDAR, barometric altitude readings, and inertial data to determine position in real time.

The framework, called CLAK, was created by a research team at Prince Sultan University. It is designed for environments where traditional navigation systems fail—such as tunnels, dense urban areas, forests, or conflict zones—where satellite signals are often blocked, disrupted, or spoofed.

Figure 1. Drone Navigation.

Unlike conventional methods that depend heavily on GPS or visual inputs, CLAK eliminates these dependencies. Visual navigation systems, while useful, require sufficient lighting, recognizable textures, and significant computational power, making them unreliable in low-visibility or resource-constrained scenarios. CLAK overcomes these limitations by learning directly from non-visual sensor data, enabling drones to maintain orientation and positioning even in challenging conditions. Figure 1 shows drone navigation.

At the heart of the system is a sophisticated AI pipeline that blends multiple techniques. Convolutional layers extract patterns from raw sensor inputs, while bidirectional LSTM networks analyze motion over time. An attention mechanism prioritizes the most relevant data, and a Kolmogorov-Arnold Network produces precise position estimates, including latitude, longitude, and elevation.

To train and validate the model, researchers used simulated flight environments built with ROS2-based tools such as Gazebo, PX4, and Q Ground Control. Real-world terrain data from the Taif region of Saudi Arabia was also incorporated to enhance realism and robustness.

Testing results were impressive. The system reduced positioning errors from more than three meters to less than one meter, achieving accuracy improvements of over 75 percent in some scenarios. Despite its advanced capabilities, the model remains lightweight, making it suitable for deployment on drones with limited onboard computing resources.

By enabling reliable navigation without GPS or cameras, this AI framework could significantly expand the operational range of drones, particularly in complex or signal-denied environments where precision and adaptability are critical.

reference:

1. https://interestingengineering.com/military/drones-no-gps-ai-navigation

Cite this article:

Keerthana S (2026), No GPS, No Cameras—Just AI: The Future of Drone Navigation, AnaTechMaz, pp.361