Research and Technical Justification of the Optimal Architecture for a Robotic Kinetic Interception Complex (C-UAS) for Protecting Armored Vehicles from FPV Drones

In the context of the modern battlefield, characterized by the massive use of cheap FPV drones, traditional means of protecting armored vehicles demonstrate insufficient effectiveness. This report presents a comprehensive study aimed at developing and technically justifying the architecture of a robotic active kinetic protection complex (Counter-Unmanned Aerial System, C-UAS).

The report sequentially examines the operational-tactical context of the threat, conducts a comparative analysis of possible weapon systems, justifies the selection of design solutions for the electromechanical platform, sensor complex, and computational architecture. Special attention is paid to software algorithms for detection, tracking, and interception of targets, as well as issues of system integration onto the platform and the economic efficiency of the solution.

The result of the work is a detailed technical and economic model of a compact, high-speed autonomous module capable of detecting, tracking, and destroying FPV drones in real-time as they approach the protected object, closing the "detection-destruction" loop without the need for crew intervention.

1.1. Battlefield Transformation and the Crisis of Traditional Protection

Modern low-intensity conflict is characterized by a shift to an asymmetric, distributed, and robotic model of warfare. FPV drones, originally civilian racing devices, have become high-precision weapons capable of delivering pinpoint strikes against critically important elements of armored vehicles (AFVs): the engine, turret, onboard electronics, and weakly armored upper projections.

1.2. Target Profile: Kinematic and Physical Characteristics of an FPV Drone

For designing an effective interception complex, a detailed analysis of the typical target profile is necessary. An FPV drone represents a difficult target to engage due to a number of characteristics:

• High Speed: 80–140 km/h (22–39 m/s).
• Ultra-low RCS: dimensions 300x300x150 mm, carbon fiber frame.
• High Maneuverability and Acceleration: capability for sharp changes in velocity vector.
• Complex Acoustic and Thermal Profile: low-power brushless motors.
• Ultra-low Flight Altitude: operation at extremely low altitudes (2–10 m), utilization of terrain folds.

Design Implication: A detection system is required that can detect a low-observable target at a range that provides a time window for platform rotation, acquisition, tracking, and firing. The total system reaction time (OODA loop) must be less than 2 seconds.

2.1. Comparative Analysis: Rifled vs Smoothbore Weapons

Analysis has shown that for the task of engaging a small, highly maneuverable target within the constraints of mass, dimensions, and recoil on a mobile platform, a smoothbore weapon chambered for 12-gauge ammunition is the optimal choice.

2.2. Analysis of 12-Gauge Ammunition Types

Examination of various types of loads for smoothbore shotguns allowed for the determination of the most effective ammunition for engaging drones.

Conclusion: A 12-gauge cartridge with #1 shot (4.0 mm) represents the optimal balance between pattern density, retained kinetic energy of each pellet at ranges up to 80 m, and suitability for use in an automated system with magazine feeding.

2.3. Weapon Platform Selection

Based on the conducted analysis, a semi-automatic carbine/shotgun chambered for 12/76 ammunition with the capability for magazine feeding (drum or box magazine of 5-10 rounds) has been selected as the base platform. Criteria: reliability of the action in various positions, compatibility with different ammunition types, presence of standard mounting interfaces (Picatinny rail) for integration with the gimbal platform.

3.1. Actuator Analysis: The Problem of Speed and Torque

The key task of mechatronics is to provide a slew rate that exceeds the angular velocity of the target. For an FPV drone attacking from a range of 100 m, the required angular velocity of the platform can reach 2 rad/s (~115°/s).

3.1.1. Stepper Motors

Advantage in positioning accuracy and control simplicity. Disadvantage: sharp drop in torque at high speeds, resonance phenomenon, and missed steps. Unsuitable for high-dynamic tracking.

3.1.2. Servo Drives and Harmonic Drives (Strain Wave Gears)

Commercial servo drives with reducers (e.g., based on Harmonic Drive) provide high torque in compact sizes. Problem: the reducer introduces backlash, reducing tracking and holding accuracy, as well as mechanical losses.

3.1.3. Direct Drive (Direct Drive Brushless Motors)

Optimal solution for high-precision and high-speed pointing tasks. Eliminating the reducer removes backlash, mechanical losses, and increases reliability. Modern frameless torque motors provide the highest dynamics, accuracy, and overload capability. Require the use of high-torque motors and sophisticated control algorithms (FOC - Field Oriented Control).

3.2. Recoil Mitigation System (Soft Recoil Mitigation)

To maintain firing accuracy in burst mode and enable rapid re-acquisition after firing, an active or passive recoil mitigation system is necessary. Proposed solution: a combination of elastic-damping elements in the weapon mounting to the platform (passive mitigation) with an algorithm for predictive compensation in the drive based on accelerometer data (active stabilization).

4.1. Edge AI Hardware Platform

Processing of video streams, execution of neural network detection and tracking algorithms, ballistic calculations, and drive control must be performed in real-time onboard the system. A platform with high performance in integer operations (INT8) and efficient power consumption is required.

Selection: The NVIDIA Jetson Orin Nano module represents the optimal balance of performance, power consumption, availability of development tools (JetPack SDK, TensorRT, DeepStream), and peripheral support.

4.2. Optical System

Primary sensor — a wide-angle (≥90° horizontally) visible spectrum camera with a global shutter (not rolling shutter) for detecting targets on approach. Resolution 1920x1200 at a minimum of 60 FPS. Additionally, for precise tracking and range determination, a narrow-angle zoom camera on a separate servo platform or a stereo camera may be used.

4.3. Radio Frequency Cueing (RF Cueing)

To reduce reaction time and detect the drone before it appears visually in the camera's field of view, the use of a passive radio direction finder is proposed. This subsystem scans the 900 MHz, 1.2 GHz, 2.4 GHz, 5.8 GHz bands (typical for FPV control and video transmission), determines the direction to the source, and provides azimuth and elevation angles to the main optical system for predictive slewing.

5.1. Detection: YOLO Neural Networks

For detecting small and fast targets, an optimized version of the YOLO (You Only Look Once) neural network architecture is used, such as YOLOv8n or a specially trained YOLO-Fastest. The model is trained on a dataset containing thousands of images of FPV drones from various angles, in different weather conditions, and against contrasting backgrounds. Inference is performed on TensorRT for maximum speed (target > 30 FPS).

5.2. Tracking and Prediction: Kalman Filter

After detection, the target is passed to a tracking algorithm. An Extended Kalman Filter (EKF) is used to estimate the target's state vector (position, velocity, acceleration in pixel and angular coordinates), filter detection noise, and predict the target's position for the shot time-of-flight.

5.3. Ballistic Calculator and Lead Calculation

Based on the predicted target coordinates, known ballistics of the selected cartridge (initial shot velocity, ballistic coefficient), and current meteorological conditions (temperature, pressure, wind, entered manually or via a weather sensor), the aiming point with lead is calculated. The calculation is performed numerically, accounting for the nonlinear velocity drop of the shot.

5.4. Control: Visual Servoing

For precise aiming, the Visual Servoing method (control via visual feedback) is applied. The algorithm continuously calculates the error between the current target position in the frame and the center of the reticle, converting this error into control voltages for the gimbal platform actuators. This allows the system not just to "chase" the target, but to compensate for its maneuvers in real-time.

6.1. Power Supply and Placement

The complex is designed as an autonomous module, mounted on the turret or hull of the AFV. Power is supplied from the vehicle's onboard network (24/28 V) via a converter. The mass of the target prototype should not exceed 50 kg to ensure acceptable load on the chassis and the possibility of mounting by the crew.

6.2. Operating Modes

• Autonomous: Full detection and engagement cycle without operator involvement. Primary mode.
• Manual (Override): The operator, via a control panel or tablet, sees the camera feed and can select a target, cancel an attack, or fire manually.
• Standby: The system conducts panoramic surveillance, detects and classifies targets, but does not open fire without command.

6.3. War Economics

Key Aspect. The cost of a single FPV kamikaze drone ranges from $300 to $1000. The cost of a 12-gauge cartridge is approximately $1-5. Even with a hit probability P_hit = 0.7, the economic suppression of an attack by a swarm of 10 drones is orders of magnitude more cost-effective than the loss of a single armored vehicle costing from $2 million. The system pays for itself from the first successfully repelled attack.