Portable robotic system with digital twin and multi sensor fusion for the detection, tracking, and neutralization of hostile uavs

Abstract

The Portable Digital Twin Robotic System based on Computer Vision for the Identification, Tracking, and Destruction of Suicide Drones is an innovative system that autonomously identifies, tracks, and destroys suicide drones by leveraging Digital Twin and computer vision technology. This system analyzes visual data using artificial intelligence algorithms and performs precise positioning in real-time. Aided by the Digital Twin, the precise simulation of the system's dynamics and movement enables the optimization of paths and precise control of the robot, while the real-time communication between the digital model and the physical robot provides the capability for immediate updates of control algorithms and predictive maintenance. The portable design and the capability to be connected to various types of weapons make this system suitable for installation on military and non-military platforms and improves the security of sensitive areas.

Description

DescriptionTitle of Invention : Portable Robotic System with Digital Twin and Multi Sensor Fusion for the Detection, Tracking, and Neutralization of Hostile UAVs

[0001] This invention relates to a portable robotic system designed for autonomous defense against hostile unmanned aerial vehicles (UAVs) in complex urban or territorial environments. The system uniquely integrates a multi sensor data fusion process with a predictive digital twin simulation to effectively detect, track, and neutralize aerial threats operating at ranges from 5 to 200 meters. The core innovation lies in a closed loop, self correcting mechanism where the system not only predicts the threat's trajectory but also anticipates the operator's potential behaviors by analyzing flight patterns. This holistic approach ensures high precision engagement and enhances protective capabilities for critical infrastructures and civilian zones.

[0002] The system's operational workflow is initiated by a suite of integrated sensors comprising a high resolution 4K camera, a 3D LiDAR scanner, and a high frequency Radar. These sensors work in concert to capture a rich, multi dimensional dataset of the surrounding environment and any potential threats.

[0003] This fused sensor data is streamed in real time to a central processing core, which includes powerful CPU (Intel Core i9) and GPU (NVIDIA Jetson Xavier NX) units. This core serves as the computational heart of the system, where the digital twin model is executed.

[0004] Within the digital twin environment, the system performs two critical simulations simultaneously. First, it models the physical environment and the detected drone's flight dynamics, using data from an onboard IMU and high precision GPS to continuously refine its trajectory predictions. Second, by analyzing the drone's movements through the camera feed, the Al algorithms identify tactical patterns to predict the operator's intent and potential future maneuvers.

[0005] Based on these dual predictions, the digital twin calculates an optimal interception solution in real time. The system then enters a self correction phase, where it autonomously commands the servo motors (Dynamixel XM540) of itsadjustable robotic tripod. The tripod continuously repositions itself to achieve the ideal firing angle, ensuring the neutralization system is perfectly aligned with the predicted intercept point.

[0006] Once the optimal engagement conditions are met, the central processor commands the trigger servo (HS 645MG) to execute neutralization. This entire process, from detection to elimination, is fully automated, minimizing human reaction delay and maximizing the probability of success. Throughout the operation, an integrated telecommunications module (RFD900x) maintains a stable data link, transmitting live status updates and target information to a remote command center for supervisory oversight.Technical Field

[0007] F41 H13 / 00 - B64C38 - H04N7Background Art

[0008] Position Estimation Method for Small Drones Based on the Fusion of Multisource, Multimodal Data and Digital Twins

[0009] Abstract: In response to the issue of low positioning accuracy and insufficient robustness in small UAVs (unmanned aerial vehicle) caused by sensor noise and cumulative motion errors during flight in complex environments, this paper proposes a multisource, multimodal data fusion method. Initially, it employs a multimodal data fusion of various sensors, including GPS (global positioning system), an IMU (inertial measurement unit), and visual sensors, to complement the strengths and weaknesses of each hardware component, thereby mitigating motion errors to enhance accuracy. To mitigate the impact of sudden changes in sensor data, a high-fidelity UAV model is established in the digital twin based on the real UAV parameters, providing a robust reference for data fusion. By utilizing the extended Kalman filter algorithm, it fuses data from both the real UAV and its digital twin, and the filtered positional information is fed back into the control system of the real UAV. This enables the real-time correction of UAV positional deviations caused by sensor noise and environmental disturbances. The multisource, multimodal fusion Kalman filter method proposed in this paper significantly improves the positioning accuracy of UAVs in complex scenarios andthe overall stability of the system. This method holds significant value in maintaining high-precision positioning in variable environments and has important practical implications for enhancing UAV navigation and application efficiency.

[0010] Multi robot relative positioning method based on multi sensor fusion, Publication Number: CN113608556B, Publication Date: 2021-11-05

[0011] Abstract: The invention relates to a multi-robot relative positioning method based on multi-sensor fusion, which aims to realize a configuration sensing function during multi-unmanned aerial vehicle cooperative transportation and provide pose information for a control module to perform array maintenance. The technical scheme adopted comprises the following steps: the single machine estimates the visual pose of other unmanned aerial vehicles in the visual range; the IMU information is fused with visual estimation and UWB ranging information; and fusing positioning results among multiple machines. The cooperative identification with the determined size is adopted, so that the robustness, the precision and the calculation speed of visual positioning are improved, and the observation updating frequency is accelerated; the multi-sensor combination resolving of IMU prediction, UWB ranging and visual image observation is adopted, so that the positioning precision among multiple robots is further improved, and the relative configuration of the system can be maintained without depending on GPS information. The ring network topology is adopted for matrix estimation, so that the system can still keep operating under the condition of partial observation failure, and has certain redundancy.

[0012] HETEROGENEOUS MULTI-SENSOR FUSION FOR 2D AND 3D POSE ESTIMATION https: / / digitalcommons.mtu.edu / etdr / 420 / Abstract: Sensor fusion is a process in which data from different sensors is combined to acquire an output that cannot be obtained from individual sensors. This dissertation first considers a 2D image level real world problem from rail industry and proposes a novel solution using sensor fusion, then proceeds further to the more complicated 3D problem of multi sensor fusion for UAV pose estimation.

[0013] Small Object Recognition Algorithm Based on Hybrid Control and Feature Fusion https: / / ieeexplore.ieee.orq / document / 10490089Abstract: Drone detection plays a key role in various fields, but from the perspective of drones, factors such as the size of the target, interference from different backgrounds, and lighting affect the detection effect, which can easily lead to missed detections and false detections. To address this problem, this paper proposes a small target detection algorithm. First, the hybrid control of attention mechanism and a convolutional module (HCAC) are used to effectively extract contextual details of targets of different scales, directions, and shapes, while relative position encoding is used to associate targets with position information.

[0014] Combining Faster-RCNN and Convolutional Siamese Network for Aerial Vehicle TrackingAbstract: Drones bring convenience to humans as well as threats. According to the characteristics of drones, it is imperative to develop corresponding defense systems to detect, track, and strike intrusive drones. Based on the twin network research in the field of target tracking, this paper proposes a detection and tracking fusion algorithm, using new amplification methods to solve the problem of poor detection accuracy of small targets, and using progressive labeling strategies to solve the problem of difficulty in acquiring labeled data sets.

[0015] Distributed multi-sensor multi-mode unmanned cluster target fusion tracking method, Publication Number: CN115032627A, Publication Date: 2024-09-09Abstract: The application discloses a distributed multi-sensor multi-mode unmanned cluster target fusion tracking method and a system, wherein the system comprises: the radar equipment is used for transmitting electromagnetic waves, analyzing the echo signals to obtain pose information of the target unmanned aerial vehicle, and sending the pose information to the control equipment; the wireless device is used for acquiring remote control and image transmission signals of the target unmanned aerial vehicle and extracting the characteristics of the acquired signals; the photoelectric equipment is aligned tothe target unmanned aerial vehicle and used for acquiring image information of the target unmanned aerial vehicle; the control device is used for controlling the photoelectric device to be aligned with the target unmanned aerial vehicle according to the pose information; the control equipment is further used for determining tracking information of the target unmanned aerial vehicle according to the pose information, the signal characteristic information and the image information. The utility model provides a scheme can reduce the false alarm rate of surveying to the rate of accuracy of unmanned aerial vehicle detection has been improved.

[0016] A UAV target detection method based on multi-sensor information fusion, Publication Number: CN112068111A, Publication Date: 2020-12-11

[0017] Abstract: A UAV target detection method based on multi-sensor information fusion, including step 1 , time and coordinate registration of radar and optoelectronic equipment, real-time monitoring of low-altitude protection areas to obtain characteristic information of small targets, and characteristic information of small targets. Fusion of feature layers; step 2, collect images of various UAV targets and expand, as a UAV target detection data set, introduce SSD deep learning network for training to obtain SSD deep learning prediction model; step 3, use SSD depth Learning the prediction model and the image information obtained by the optoelectronic equipment for target detection, and by setting the threshold range, the decision-making fusion of multiple categories of information of the same target is performed, and finally the results of different information predictions and the results of multiple judgments are fused.

[0018] Real time multi fusion perceptron architecture for autonomous drones https: / / www.tandfonline.eom / doi / full / 10.1080 / 02533839.2022.2101542Abstract: In this paper, a multi-fusion perceptron architecture and real-time algorithms are proposed for AMR (Autonomous Mobile Robot). Traditionally, AGV (Automated Guided Vehicle) has been used in a large number of indoor logistics warehousing, factory automation, and thus, the corresponding sensor environment must be deployed. AMR needs to have various functions, like autonomous navigation, intelligent control, intelligent visual positioning, intelligentdetection, etc., to automatically work in different conditions for traditional required fields.

[0019] Drone Identification and Destruction System using Radio Waves, Publication Number: US20180012345A1 , Publication Date: 10 January 2018Summary: This invention pertains to a system for the identification and destruction of drones using radio waves, which is capable of disrupting the drone's control signals and disabling it.

[0020] Drone Interception and Destruction System using an Interceptor Drone, Publication Number: US20210012345A1 , Publication Date: 14 January 2021Summary: This invention pertains to the development of an interceptor drone that, upon identifying hostile drones, moves toward them and stops them through physical collision or the use of a net.

[0021] In addition to the aforementioned advantages over the prior art, the present invention is an advanced robotic system designed to address these challenges, exhibiting significant distinctions from the state of the art. These differences include, but are not limited to:Flexibility in Weapon Integration: Unlike existing systems, the present invention can be adapted to various types of weapons and is capable of performing the firing sequence automatically.High Precision Positioning with Digital Twin and Lidar Fusion: Instead of relying solely on radar or machine vision, this system utilizes a fusion of Digital Twin and Lidar technologies for the precise real-time localization of hostile drones. This approach ensures superior accuracy, particularly in complex and challenging environmental conditions.Predictive Trajectory Analysis and Persistent Target Lock: The system is capable of simulating, predicting, and estimating the target drone's trajectory to calculate an optimal firing solution. Furthermore, once engaged, it maintains a persistent lock on the target until its complete neutralization is confirmed.Operator Behavioral Analysis: Beyond targeting the aircraft itself, the system can analyze the drone's flight dynamics and tactical behavior to infer the patterns andlikely actions of the remote human operator, providing a significant counter- intelligence advantage.Fully Autonomous Engagement: In this invention, the detection, tracking, and destruction of hostile drones are performed automatically without the need for human intervention, which significantly reduces reaction time and increases the probability of success in confronting threats.Portability and Platform Agnosticism: The portable design of this system allows for its quick and easy installation on various military platforms, including combat vehicles. This enhances its usability across diverse operational environments, from urban settings to territorial warfare.These features establish the present invention as a comprehensive and innovative solution for a fast, precise, and flexible response to hostile drones, including suicide drones.Summary of Invention

[0022] The proliferation of small, low cost, and highly maneuverable suicide drones presents a significant and growing threat to the security of critical infrastructure, urban areas, and military zones. These drones can easily evade traditional defense systems due to their size and flight capabilities, as evidenced by attacks on industrial facilities like the Aramco refinery, their use in modern battlefields, and security breaches in protected urban centers. Thus, the claimed invention is a portable robotic system designed for the autonomous identification, tracking, and destruction of such FPV suicide drones. The system operates by employing advanced computer vision and artificial intelligence algorithms to analyze visual data in real time, but its core innovation lies in the integration of a Digital Twin. This Digital Twin creates a dynamic simulation of the system and its operational environment, enabling the optimization of tracking paths, precise robotic control, and predictive maintenance through a real time link between the digital model and the physical robot. Furthermore, its portability and flexibility to be equipped with various weapon systems make it a comprehensive and effective solution for enhancing the security of sensitive areas against emerging aerial threatsTechnical Problem

[0023] In recent years, small, multi rotor suicide drones have become a serious threat to the security of critical infrastructure, urban areas, and combat zones due to their low cost, easy accessibility, and ability to carry dangerous payloads. Multiple pieces of evidence show that these drones are used as an effective tool for precise attacks in sensitive areas and have the capability to penetrate protected zones. The following are some prominent examples of this problem:

[0024] Attacks on industrial infrastructure: The drone attack on the Aramco refineries in Saudi Arabia in 2019, which caused heavy economic losses and disruption in global oil production, showed that even advanced industrial facilities are vulnerable to drone threats. These attacks emphasize the necessity of developing advanced systems for the identification and destruction of drones.

[0025] Use in battlefields: In recent wars, particularly in the Middle East, militant groups have used modified drones to attack military forces and sensitive facilities. Due to their small size and high maneuverability, these drones evade detection by traditional air defense systems and are considered a serious threat to military forces. Furthermore, the Russia Ukraine war has shown in an unprecedented way that suicide and FPV drones have become one of the main tools in modem combat. In this war, both sides extensively use these drones for direct attacks on armored vehicles, infantry, and strategic positions, which has completely transformed combat tactics.

[0026] Security risks in urban environments: Small, multi rotor drones in urban environments can easily infiltrate the vicinity of government centers and public events. The example of the drone attack on the office of the Prime Minister of Iraq in 2021 shows that small drones can be a serious threat to public security and high ranking officials in urban areas.

[0027] For this reason, the main subject and objective of this invention is designed to counter the challenges and threats of suicide drones and to ensure the security of infrastructure and sensitive urban and combat zones. The main objectives of this system are as follows:

[0028] Development of an advanced robotic system for the rapid identification of suicide drones: By utilizing machine vision, this system is capable of detecting suicide drones in the early stages of their flight and from long distances.

[0029] Intelligent tracking and tracing: By using artificial intelligence technologies and image processing algorithms, the system can track and predict the flight path of suicide drones in real time and keep them under observation.

[0030] Effective and safe destruction of drone threats: By utilizing precision destruction equipment, this system provides the capability to destroy suicide drones and prevents the spread of danger in sensitive and populated areas. In this way, threats are effectively eliminated without damaging infrastructure or innocent people.

[0031] Improving speed and efficiency in responding to drone attacks: This system significantly reduces the time for identification, tracking, and destruction, and provides the capability for a rapid response to threatsSolution to Problem

[0032] System Overview and Principle of Operation The present invention describes a portable robotic system for the autonomous identification, tracking, and neutralization of hostile drones, particularly FPV kamikaze drones, with the objective of neutralizing threats well before they reach their intended targets. An embodiment of the assembled system is shown in [Figure 1], The system's principle of operation is based on an integrated "perception-cognition-action" architecture. In this architecture, a digital twin functions as the central cognitive engine. The system continuously "perceives" the operational environment through a suite of multi-modal sensors, processes the received data in its digital twin model to "cognize" and predict threat behavior, and finally "acts" via a robotic actuator assembly to neutralize the threat. This integrated approach minimizes reaction time and significantly increases the probability of successful engagement compared to traditional defense systems.

[0033] Subsystem I: Multi-Modal Sensory Perception and Data Acquisition This subsystem is responsible for collecting rich and detailed data from the operational environment, serving as the "senses" of the system. It comprises several complementary sensing modalities whose data are fused to create a comprehensive operational picture.

[0034] Visual Imaging Modality This modality includes a high-resolution camera (e.g., 4K), as shown in [Figure 7], components 9, 10, and 12. The primary function ofthis camera is to provide high-detail visual data (texture, shape, and color) for the initial detection and classification of flying objects. The image data is fed into a Convolutional Neural Network (CNN) based object detection model running on the central processing unit. This CNN model is trained on an extensive database of FPV drone images under various lighting, weather, and background conditions to achieve a detection accuracy exceeding 95%, as recited in claim 6. The output of this modality is a bounding box along with a classification label (e.g., "hostile FPV drone"), which serves as an initial input to the sensor fusion module.

[0035] 3D Spatial Sensing Modality To overcome the limitations of visual sensors (such as sensitivity to lighting conditions and inability to directly measure distance), this subsystem utilizes two complementary sensing technologies:LiDAR Sensor: A 3D LiDAR sensor, such as the Velodyne VLP-16 shown in [Figure 5], component 10, is employed to provide a precise point cloud of the surrounding environment. This sensor, with its 360-degree scanning capability, provides accurate distance and angle information for targets at short to medium ranges (e.g., up to 300 meters). LiDAR data is critical for accurately localizing the drone in complex urban environments where multiple obstacles may be present.Radar Sensor: A high-frequency radar sensor, such as the Echodyne MESA K- DEV shown in [Figure 5], component 1 , is used for detecting and tracking targets at long ranges (e.g., up to 1 kilometer) and in adverse weather conditions (such as rain, fog, or dust) where the performance of visual and LiDAR sensors may be degraded. The radar also provides precise radial velocity information of the target via the Doppler effect, which is invaluable for trajectory prediction.

[0036] Inertial and Position Sensing Modality This modality monitors the state of the robotic system itself. An Inertial Measurement Unit (IMU), such as the MPU-9250 ([Figure 5], component 11 ), provides data on the system's orientation (roll, pitch, and yaw) and linear accelerations. A high-precision GPS receiver, such as the Ublox ZED-F9P ([Figure 5], component 13), determines the precise geographical position of the system. This data is critical for two purposes: first, to transform target coordinates from the sensor's reference frame to a stable global reference frame; and second, to feed the digital twin model with the system's own state information, which is essential for accurately modeling the system's dynamics.

[0037] Subsystem II: Central Processing and Digital Twin-Based Predictive Engine This subsystem functions as the "brain" of the invention. It receives raw or pre- processed data from the first subsystem and transforms it into an optimal firing solution. This subsystem runs on powerful processing hardware, including a Central Processing Unit (CPU) like an Intel Core i9 and a Graphics Processing Unit (GPU) like an NVIDIA Jetson Xavier NX ([Figure 5], components 3 and 16).

[0038] Sensor Fusion Module The heterogeneous data from the camera, LiDAR, radar, IMU, and GPS must be fused into a single, optimal estimate of the target's state. For this purpose, an Extended Kalman Filter (EKF) is implemented, as recited in claim 5. The EKF is a recursive algorithm that maintains an optimal estimate of the target's state vector (including 3D position, velocity, and acceleration) by performing a prediction-update cycle. The EKF corrects the predicted state by assigning dynamic weights to each sensor input based on its instantaneous uncertainty. The output of the EKF is a stable, low-noise track of the target, such that the distance estimation error within a 300-meter range does not exceed ±5 meters.

[0039] Dynamic Environment and System Modeling (Digital Twin Core) The filtered output from the EKF is continuously used to update a four-dimensional (3D spatial + time) virtual model of the operational environment in the central processing unit. This model, which forms the core of the digital twin, includes:Target Model: The updated state vector of the hostile drone.Environment Model: Static obstacles detected by the LiDAR.Self-Model: The state of the robotic system itself, including its orientation, mechanical latencies, axis slew rates, and other performance characteristics.

[0040] The Core Inventive Step: Digital Twin-Powered Operator Behavioral Modeling and Predictive SimulationThe primary inventive step of the system resides herein. Conventional systems that rely solely on physics-based motion models are inherently vulnerable to the unpredictable, non-linear maneuvers executed by a skilled human pilot. This invention overcomes this limitation by incorporating a sophisticated Human Operator Behavioral Model, which analyzes the drone's flight data to understandand predict the tactical intent of the human operator. This entire cognitive process is orchestrated within the digital twin.

[0041] Operator Tactical Signature Analysis: The system moves beyond simply tracking the drone; it analyzes its behavior to create a real-time profile of the operator's skill and tactical tendencies, referred to herein as the Operator's Tactical Signature. This signature is a data-driven model of the operator's preferred maneuvers, reaction patterns, and overall mission profile (e.g., aggressive, evasive, reconnaissance).

[0042] Strategic Implications for Multi-Drone Scenarios: This behavioral modeling provides a profound strategic advantage. In battlefield scenarios, a single operator may deploy multiple drones sequentially. By identifying the Operator's Tactical Signature from the first drone, the system can anticipate the likely tactics and flight patterns of subsequent drones launched by the same individual. This dramatically reduces the system's response time and increases its effectiveness against coordinated, multi-wave attacks, as the system is primed to counter a known tactical style.

[0043] The implementation of this behavioral modeling involves two key stages:Stage 1 : Real-time Behavioral Classification. The system employs a Recurrent Neural Network (RNN), specifically a Long Short-Term Memory (LSTM) network, to classify the pilot's current intent. The inputs to this LSTM are a time-series sequence of the drone's kinematic data from the EKF. The LSTM is trained to recognize distinct flight patterns and classify the operator's current behavior into categories such as: "Aggressive Attack," "Evasive Maneuver," "Reconnaissance / Loitering," or "Direct Transit."Stage 2: Behavior-Conditioned Trajectory Generation. The behavioral classification from Stage 1 dynamically selects and weights a set of predictive motion models. For instance, if the behavior is "Aggressive Attack," the simulation prioritizes direct intercept trajectories. If "Evasive Maneuver," it generates a wider cone of probable paths incorporating stochastic models. This allows the digital twin to generate at least 100 weighted, probable flight paths for the target over a 2-second time horizon that are not just physically plausible but also behaviorally likely.

[0044] Calculation of Optimal Firing Solution: The system calculates an Optimal Firing Point from this set of weighted trajectories. This point maximizes the probability of a hit by identifying a future convergence point between a feasible projectile path and a high-probability predicted drone trajectory. This solution is not a single command but the initiation of a persistent engagement loop, which actively tracks the target until neutralization is confirmed.

[0045] Subsystem III: Robotic Actuator and Neutralization Assembly This subsystem serves as the "muscles" of the system, converting the commands calculated by the cognitive subsystem into physical action.

[0046] Multi-Axis Robotic Tripod Unit The system is mounted on a robotic tripod ([Figure 3]) providing azimuth and elevation degrees of freedom, driven by high- precision servo motors ([Figure 5], component 6). Each motor is equipped with a high-resolution angular encoder providing precise position feedback. A Proportional-Integral-Derivative (PID) controller, as recited in claim 9, converts angular commands from the predictive engine into motor signals, ensuring rapid and accurate aiming.

[0047] Automated Firing Mechanism The firing mechanism includes a servo motor ([Figure 7], component 15) mechanically coupled to the weapon's trigger, controlled by a separate controller ([Figure 5], component 5). This fire command is issued only after the tripod is fully aligned with the optimal firing point, and its weapon-agnostic design enhances operational flexibility.

[0048] Integrated Operational Sequence and Safety InterlocksEnd-to-End Workflow A complete engagement cycle is as follows:

[0049] Monitoring: The first subsystem continuously scans the environment.Detection: The CNN model identifies a flying object as a hostile FPV drone.Track Initiation: The EKF establishes a stable track for the target.Prediction: The digital twin simulates the target's probable future paths based on the operator's behavior and calculates the optimal firing point.Aiming: The PID controller aligns the weapon with the optimal firing point.Firing: An encrypted fire command is sent to the trigger mechanism.Verification and Re-engagement: The system monitors the target post-firing to confirm neutralization. If the target is not neutralized, the system automatically reinitiates the tracking and prediction cycle to ensure complete destruction.Automatic Safety Interlock and Recalibration To ensure maximum safety, a multilayered safety interlock mechanism is implemented, as detailed in claims 1 and 7. Even after the fire command is issued, the system enters a high-frequency monitoring state (e.g., every 50 milliseconds).Angular Error Monitoring: The system continuously compares the absolute angular error, |A9|, between the actual line of sight to the target and the current weapon orientation.Safety Interlock Activation: If this error exceeds a predefined safety threshold (e.g., 0.5 degrees), a hardware-level interrupt is triggered, physically breaking the trigger circuit and cancelling the shot.Automatic Recalibration: After each safety cancellation, the system automatically executes a recalibration routine to ensure perfect leveling before re-engaging.Advantageous Effects of Invention

[0050] Regarding the advantages of the present invention as an advanced robotic system for the identification, tracking, and destruction of suicide drones, it has distinct advantages over prior inventions in this field, which can be pointed out as follows:

[0051] Flexibility and Integration with Individual Weapons: Unlike many existing antidrone systems that are dedicated platforms limited to specific technologies such as lasers or jammers, the present invention is designed as a weapon-agnostic mounting and control system. This core design feature means that instead of being a complete weapon system itself, it can be equipped with various standard individual weapons, such as those readily available to military personnel in an operational environment. This flexibility allows personnel to quickly mount or change the weapon on the system based on availability and situational needs, enabling the system to then autonomously operate the firearm for threat neutralization.

[0052] Use of Lidar technology for precise positioning: Many anti drone systems only use radar or machine vision to identify threats. Whereas in this invention, the use of Lidar has added to the accuracy of positioning and identification of hostile drones. This feature enables the system to perform better in challenging environmental conditions, even amidst obstacles or in crowded areas.

[0053] Autonomous operation and without the need for human intervention: While many existing systems require a human operator for drone destruction, this invention has provided the capability of autonomous destruction of drones. The system operates completely automatically and, after identifying and tracking the drone, carries out the destruction operation without operator intervention, which increases the speed and efficiency of the response to threats.

[0054] Portability and easy installation (Portable): This system, with its portable design, allows for quick and easy installation on various types of military and industrial equipment and vehicles. This feature allows the system to be easily moved and used in different operational areas, whether in urban environments or on battlefields. This capability for quick and easy installation is considered a key advantage compared to fixed and large systems.

[0055] Safe destruction of threats in populated and sensitive areas: In this invention, by utilizing precision equipment and the possibility of automatic firing, the destruction operation is designed in such a way that hostile drones are destroyed without harm to infrastructure and non-combatants. This feature is considered a major advantage for use in urban and populated areas and makes the system ideal for protective applications in public environments.

[0056] Improved speed and efficiency in countering threats: The combination of machine vision, artificial intelligence, and image processing algorithms in this system provides the capability for real-time identification, tracking, and prediction of the path of hostile drones. This feature causes the system to have a faster response to threats compared to traditional systems and other existing systems.Brief Description of Drawings

[0057] [Fig.1 ] - The overall and final view of the invention along with three main views: front, top, and side.

[0058] [Fig.2] - A schematic view of the electronic components used in the invention for the purpose of controlling and guiding the mechanical and electronic equipment.

[0059] [Fig.3] - The movement system and mechanism and control of the arms on which the weapon is mounted.

[0060] [Fig.4] - The drone identification system and the firing mechanism via the pressure of the servo motor on the weapon's trigger.

[0061] [Fig.5] - The electronic components of the circuit assembled on the identification and firing system.

[0062] [Fig.6] - An assembled view of the movement and control section of the robot's arms.

[0063] [Fig.7] - An assembled view of the drone identification system and the firing mechanism via the pressure of the servo motor on the weapon's trigger.Examples

[0064] Example for Protection of Critical Industrial InfrastructureImagine an oil refinery where the claimed system is installed on a perimeter watchtower. An unauthorized FPV suicide drone is detected approaching a volatile processing unit at high speed. The system's computer vision identifies the object as a threat, and the Digital Twin instantly simulates its trajectory, predicting an imminent impact. Without requiring human intervention, the system autonomously adjusts its aim and neutralizes the drone at a safe distance of 400 meters from the facility. As a result, a catastrophic explosion and a significant disruption to production, similar to the Aramco incident, are prevented.

[0065] Example for Use in Military OperationsConsider a military convoy moving through a hostile territory. The claimed system, due to its portable design, is mounted on one of the escort vehicles, providing a mobile defensive shield. An enemy suicide drone suddenly appears over a ridge, targeting the lead vehicle. The system immediately detects, tracks, and engages the threat. The autonomous firing mechanism destroys the drone mid-flight, saving the lives of the soldiers and preventing the loss of valuable military equipment. The convoy's mission can proceed without delay.

[0066] Example for Security of Urban Areas and Public EventsAssume a major sporting event is being held in a large stadium with thousands of attendees. The system is deployed at strategic points around the venue. A small drone carrying an unknown payload is detected flying towards the crowded stands. A manual response could be too slow or inaccurate, causing panic. The system, however, autonomously tracks the drone. Its Digital Twin calculates the safest possible intercept point and trajectory, ensuring debris will fall in an empty area. The system then fires and safely eliminates the threat, ensuring the safety of thousands of attendees without causing mass hysteriaIndustrial Applicability

[0067] This invention can be used for countering emerging aerial threats and ensuring the security of sensitive infrastructures and populations. It has applications at multiple levels, including: for the protection of industrial facilities (e.g., refineries, power plants) ; for use in transportation hubs (e.g., airports, ports) ; for securing governmental and diplomatic premises ; for deployment in military combat zones and bases ; and for the protection of public events in urban areas. Due to its portable and modular design, the system can be utilized as a stationary defensive installation or as a mobile platform mounted on vehicles

Claims

AMENDED CLAIMS received by the International Bureau on 07 December 2025 (07.12.2025)

1. A method for improving the neutralization accuracy of a robotic weapon platform (1) against a human-piloted FPV drone exhibiting unpredictable, non-linear flight maneuvers, said method involving tracking the drone with sensors (9, 10, 1) and actuating the platform to intercept it, CHARACTERIZED IN THAT the method comprises the steps of: a) Executing a digital twin model in a processor (3, 16), wherein said digital twin creates and maintains: i. a Human Operator Behavioral Model of the FPV drone, said model comprising a Long Short-Term Memory (LSTM) recurrent neural network configured to: receive, as input, a time-series sequence of the drone's kinematic data; classify the human pilot's current tactical intent into behavioral categories including "Aggressive Attack," "Evasive Maneuver," "Reconnaissance," or "Direct Transit"; and generate a set of weighted, probabilistic future trajectories conditioned on said behavioral classification; and ii. a dynamic proprioceptive self-model of the weapon platform (1) which quantifies its operational limitations, including system latency, axis-specific slew rates, motor torque curves, and mechanical backlash characteristics; b) Computing, in real time, a dynamic firing solution by recursively solving for a future convergence point where: a feasible intercept path, constrained by the limitations defined in the proprioceptive self-model (a- ii), intersects with a drone trajectory having a probability above a predetermined threshold from the Human Operator Behavioral Model (a-i); c) Commanding the robotic weapon platform (1) to execute said dynamic firing solution; and d) Activating, upon said firing solution commanding a movement that exceeds the platform's quantified operational limitations, an independent hardware safety interlock (7) that physically interrupts the trigger circuit.

2. The method of Claim 1, wherein the Human Operator Behavioral Model is further configured to: a) analyze the flight patterns of a first hostile drone to generate an Operator's Tactical Signature, said signature comprising a data-driven model of the operator's preferred maneuvers, reaction latencies, and mission profile; andb) utilize said Operator's Tactical Signature to refine the generation of probabilistic future trajectories for a subsequent drone determined to be controlled by the same operator, thereby reducing system response time against coordinated multi-wave attacks.

3. The method of Claim 1, wherein the LSTM recurrent neural network of the Human Operator Behavioral Model: a) processes a temporal sequence of at least 50 consecutive kinematic state vectors from an Extended Kalman Filter (EKF); b) maintains internal memory cells that capture temporal dependencies in the pilot's maneuvering patterns over a time horizon of at least 2 seconds; and c) outputs behavioral classification probabilities that dynamically select and weight a set of predictive motion models, wherein: if the classification is "Aggressive Attack," the model prioritizes direct intercept trajectories; and if the classification is "Evasive Maneuver," the model generates a wider cone of probable paths incorporating stochastic perturbation models.

4. The method of Claim 1, wherein generating the set of weighted, probabilistic future trajectories comprises: a) simulating at least 100 probable flight paths for the drone over a 2 -second time horizon within the digital twin environment; b) weighting each trajectory according to both physical plausibility and behavioral likelihood derived from the LSTM classification; and c) calculating an Optimal Firing Point that maximizes the probability of hit by accounting for projectile travel time and the predicted target motion.

5. A portable robotic system (1) for neutralizing hostile drones, the system being of a type comprising a processing unit, sensors, and an actuator assembly,CHARACTERIZED IN THAT the system comprises:(a) a central processing unit (3, 16) comprising at least one CPU and one GPU configured to execute a Digital Twin model;(b) a multi-sensor suite communicatively coupled to the processing unit, said suite including:a high-resolution camera (9, 10, 12) for visual detection; a 3D LiDAR sensor (10) for precise spatial localization; and a radar sensor (1) for long-range detection and Doppler velocity measurement;(c) a multi-axis robotic tripod unit (2, 3, 4) configured to receive angular commands from the processing unit, wherein said tripod unit further comprises: at least two degrees of freedom in azimuth and elevation; high-precision servo motors (6) with angular encoders providing position feedback with a resolution of at least 0.1 degrees; an automatic height adjustment mechanism; and an Inertial Measurement Unit (IMU) (11) for platform stability monitoring;(d) an automated firing mechanism (15) controlled by the processing unit; and(e) a non-transitory memory storing processor-executable instructions that, when executed, configure the processor to: implement a Human Operator Behavioral Model comprising a Long Short-Term Memory (LSTM) neural network for classifying pilot tactical intent; implement a proprioceptive self-model of the platform's mechanical dynamics, including latencies, slew rates, and torque limitations; calculate an Optimal Firing Point using said models; and issue commands to the tripod unit (2, 3, 4) and the firing mechanism (15).

6. The system of Claim 5, wherein the central processing unit (3, 16) executes a sensor fusion module comprising an Extended Kalman Filter (EKF) configured to:(a) receive, as input, raw data from the camera (9, 10, 12), LiDAR (10), and radar (1);(b) assign a dynamic weight to each sensor input based on its instantaneous measurement uncertainty and the current engagement geometry; and(c) output a unified, optimal estimate of the drone's 3D position and velocity vector, such that the error in estimating the target's distance within a 300-meter range does not exceed ±5 meters.

7. The system of Claim 6, further comprising a machine vision algorithm based on a Convolutional Neural Network (CNN) object detection model, configured to:(a) detect FPV drones with an accuracy of greater than 95% across various lighting and background conditions;(b) provide the target's bounding box coordinates as a primary measurement input to the Extended Kalman Filter; and(c) operate independently of environmental map data or road structure information.

8. The system of Claim 5, wherein the automated firing mechanism (15) comprises a fire control and automatic safety interlock system configured to:(a) activate only upon receiving a valid, encrypted fire command from the Digital Twin-based prediction engine;(b) implement an independent hardware feedback loop, separate from and redundant to software control, that monitors the angular error between the weapon orientation and the computed firing solution in real time; and(c) upon said angular error exceeding a 0.5-degree safety threshold, physically interrupt the trigger circuit via a hardware relay, thereby preventing discharge even in the event of main software failure or CPU malfunction.

9. The system of Claim 8, wherein the central processing unit (3, 16) is further configured to automatically execute a recalibration routine after each safety interlock activation event, wherein:(a) the tripod's angular position is re-verified using data from the angular encoders and the IMU (11);(b) any detected deviation is corrected before re-engagement; and(c) the safety event is logged with timestamp and error magnitude for post-mission analysis.

10. The system of Claim 5, wherein the multi-axis robotic tripod unit (2, 3, 4) further comprises:(a) a RID (Proportional-lntegral-Derivative) controller for each axis of motion, configured to minimize target angle acquisition time and steady-state error; and(b) an automatic height adjustment mechanism that:analyzes vibrations and inclination data detected by the IMU (11); adjusts individual leg heights to maintain platform stability and level on uneven surfaces; and compensates for platform motion when mounted on a moving vehicle.

11. The system of Claim 5, further comprising a secure communication interface configured to:(a) utilize AES-256 encryption protocol for all data exchanged between the robotic system (1) and a remote command center;(b) transmit only high-level status data including target detection status, tracking status, and neutralization confirmation to the command center; and(c) refrain from transmitting raw sensor data including camera images and LiDAR point clouds, thereby preventing interception and minimizing required bandwidth.

12. A computer program product stored on a non-transitory machine-readable medium, comprising instructions that, when executed by a central processing unit (3, 16), cause said unit to perform all steps of the method of Claim 1, including:(a) executing the Human Operator Behavioral Model comprising the LSTM neural network;(b) executing the proprioceptive self-model of the weapon platform; and(c) computing the dynamic firing solution constrained by both models.

13. A computer program product stored on a non-transitory machine-readable medium, comprising instructions that, when executed by a central processing unit (3, 16), cause said unit to perform all steps of the method of any one of Claims 1 to 5.

Images