1. Introduction

Unmanned Aerial Vehicles (UAVs), commonly referred to as drones, have revolutionized various sectors, including military, surveillance, environmental monitoring, and aerial photography. One of the critical components that enhance the functionality of UAVs is the electro-optical (EO) pod, which houses cameras, sensors, and other optical systems used for imaging, targeting, and reconnaissance. The stability and precision of these systems are crucial, particularly in dynamic environments where the UAV may experience rapid movements and external disturbances such as wind or vibrations.

The effectiveness of EO pods largely depends on the control algorithms that stabilize the camera's line of sight (LOS). These algorithms must be capable of compensating for various disturbances to maintain a stable and clear image, which is essential for the success of missions ranging from military operations to commercial applications like filmmaking. Over the years, control techniques have evolved from simple Proportional-Integral-Derivative (PID) controllers to more sophisticated methods, including Kalman filters, machine learning approaches, and hybrid systems that combine multiple algorithms.

This paper focuses on the review of self-stabilization algorithms used in UAV EO pods. It examines traditional control techniques, advanced filtering methods, and the integration of machine learning to enhance stability and responsiveness. Additionally, it explores the application of these algorithms in different scenarios, highlighting their significance in both military and civilian domains. Through a comprehensive analysis of various control strategies, this paper aims to provide insights into the optimization of EO pod performance in UAVs.

2. System functionality and composition

2.1. Functionality

The functionality of UAV EO pods is to maintain a stable LOS in highly dynamic and unpredictable environments. Whether the UAV is banking, climbing, or diving, the gimbal must keep the optics oriented correctly. Additionally, it is common to encounter external disturbances that can interfere with camera focusing. Therefore, the control system must keep the UAV sensors target-locked, regardless of environmental conditions [1].

Beyond stabilization, these systems also perform tasks such as target tracking, image processing, and data transmission. The control system must be capable of following moving objects with high precision, especially in challenging environments like rapidly moving targets or cluttered areas. Additionally, the system must be resilient to environmental factors such as wind, which could otherwise induce disturbances and degrade image quality or disrupt target tracking [2].

2.2. Components

Although UAV EO pods have various types with different components and structures, they all follow similar principles. Generally, the EO pod consists of six main parts: mechanical structure, hardware, software, drive components, angle measurement components, and EO loads, all playing vital roles in maintaining the overall system [3].

The essential components of the system include optical sensors, such as high-resolution cameras that capture visual and thermal data crucial for mission success; movement and direction sensors; and the control system. A combination of motors and actuators driven by the control system performs the commands given to the gimbal assembly. The motors handle significant adjustments required to keep the gimbal aligned with the LOS of the sensor pod, while actuators, such as voice coil motors (VCMs), manage finer adjustments. These adjustments are driven by the control system's commands to maintain the angle of the sensor pod, resisting external disturbances like wind gusts or UAV maneuvers that could affect the sensor's LOS [4].

To utilize and control all components correctly, the control system is built on intricate algorithms. It takes input data from the pod's gyroscopes and sensors and sends commands to the motors and actuators. The control system's architecture is geared toward two related objectives: real-time responsiveness (making necessary adjustments promptly to keep the system stable) and accuracy (issuing commands that ensure the pod's LOS remains steady even under physical disturbances). Both objectives are critical to the pod's overall performance and, therefore, to that of the UAV itself [5].

3. Control techniques for stabilization

3.1. Traditional control techniques

3.1.1. PID control. PID controllers are fundamental tools used in many control systems due to their simplicity and effectiveness. In UAV EO pods, PID controllers maintain system stability by continuously comparing the desired position (setpoint) with the actual position. If a difference is detected, the PID controller adjusts the system by combining proportional, integral, and derivative actions. Despite their widespread use, traditional PID controllers have limitations, particularly in dynamic and non-linear environments like those encountered by UAVs. Their fixed settings make it challenging to maintain stability during sudden changes or disturbances [6].

To address these issues, improvements such as integral separation techniques have been developed. This method adjusts the integral action based on the system's deviation from the setpoint, reducing the risk of overshoot and improving stabilization in fast-moving UAVs [6]. However, PID controllers remain limited in their ability to adapt to highly variable conditions, necessitating the exploration of more advanced control methods.

3.1.2. Coarse–fine composite control. Two levels of control provide both high precision and rapid responsiveness in stabilization systems. Coarse control is managed by devices like ultrasonic motors that provide quick, large-scale adjustments, while fine control is handled by devices like voice coil motors that perform detailed, small-scale corrections to maintain system stability. The advantage of using coarse-fine composite control is that it balances speed and precision, allowing the system to respond effectively to significant disturbances while maintaining accurate positioning [7].

However, coordinating the two control levels is critical to avoid system conflicts and ensure smooth operation. This approach also requires more hardware, increasing system weight and power consumption, which must be considered during the design process [7].

3.2. Advanced control techniques

3.2.1. Filter techniques. Filter techniques are crucial for the accurate stabilization of UAV EO pods. Kalman filters are widely used for sensor fusion, effectively handling real-time data where sensor inputs may be noisy or incomplete, a common scenario in UAV operations. The Kalman filter works by predicting the future state of the system based on past measurements and updating this prediction as new data becomes available. This continuous refinement provides accurate estimates of the system's state, maintaining a stable LOS even in the presence of disturbances [8].

More advanced filters, such as Particle Filters, can handle non-linear systems and non-Gaussian noise more effectively by using a large number of particles to represent possible system states (Cai et al., 2023). These techniques are particularly effective in complex environments frequently encountered by drones. However, they also demand significant computational resources, which can be a limitation in real-time applications.

3.2.2. Machine learning approaches. A significant breakthrough in UAV stabilization has come from machine learning (ML). Reinforcement Learning (RL) and Neural Networks (NNs) can optimize control systems by learning from extensive historical and real-time data generated during UAV operations. These techniques can identify patterns and make decisions that enhance system stability and responsiveness [5].

Machine learning also plays a crucial role in predictive control. Algorithms can predict future disturbances based on past data, allowing the system to proactively adjust and maintain stability. For instance, Shen and Cheng demonstrated the use of ML-based predictive models to compensate for time-delay effects in image-tracking systems, significantly enhancing system responsiveness and stability [9].

4. Algorithm fusion

4.1. PID and Kalman filters

Combining PID controllers with Kalman filters allows a control system to benefit from the quick response of the PID controller and the noise reduction capabilities of the Kalman filter. This fusion enables the system to maintain real-time stability while ensuring control actions are based on accurate and refined state estimates. The PID controller handles immediate adjustments needed to correct errors, while the Kalman filter continuously refines state estimates, improving overall accuracy and stability of the EO pod [2, 5].

4.2. Coarse–fine composite control with advanced algorithms

Coarse–fine composite control combines rapid, large-scale adjustments with precise, fine-tuned corrections. Integrating this technique with advanced algorithms like Kalman filters or machine learning can further enhance effectiveness. For example, Kalman filters can refine coarse control actions by providing accurate state estimates, while machine learning algorithms can optimize fine control parameters in real-time based on system performance. This combined approach ensures that the EO pod can quickly respond to large disturbances while maintaining high precision in stabilization efforts [2, 5].

5. Scenario-specific applications

5.1. Military applications

For detailed intelligence gathering during reconnaissance and surveillance, systems combining Kalman filtering with coarse-fine composite control have proven effective. The Kalman filter excels in noise reduction and real-time error correction, maintaining stable imagery even under extreme conditions. Coarse-fine control allows rapid, large-scale adjustments alongside precise, fine-tuned control to keep the UAV's LOS steady in various environments [5].

Precision and speed are crucial for target acquisition and guidance missions. Integrating Kalman filters with machine learning algorithms enhances predictive tracking capabilities, allowing systems to adapt in real-time to sudden movements or changes in target behavior [2].

Stable orientation is fundamental for effective data transfer in communication relay missions. Utilizing a fusion of PID controllers and Kalman filters ensures basic stabilization tasks are managed efficiently, with the Kalman filter refining control signals and compensating for disturbances to maintain continuous communication links [1].

In direct assault missions, stabilization systems must provide clear and precise target verification. Coarse-fine composite control offers the rapid response and precision required to maintain target lock even under demanding aircraft maneuvers [7].

5.2. Civil applications

5.2.1. Consumer-level. PID controllers provide reliable stabilization during typical aerial maneuvers such as Aerial Photography and Filmmaking. Integrating machine learning can further enhance adaptability by adjusting stabilization parameters based on real-time flight conditions [5]. In drone racing, where immediate and flexible reactions are required, PID controllers secure the visual feed for racers due to their simplicity and rapid response capabilities [6].

5.2.2. Industrial-level. In chaotic environments like disaster response scenarios, a blend of coarse-fine control and Kalman filters is ideal. Coarse control manages large-scale disturbances, fine control provides operational precision, and Kalman filters ensure clear, stable images during rescue operations [2]. Urban security applications benefit from machine learning-enhanced controllers that adapt to changing environments, maintaining high-resolution imagery for effective monitoring and surveillance [5]. Robust stability is essential for maritime inspections affected by sea motion. Kalman filters provide necessary stability, maintaining clear images of offshore structures like oil rigs and pipelines despite environmental instability [8]. During Environmental, Infrastructure, and Agricultural Inspection, Stable and precise operation is critical for monitoring and inspection tasks. Machine learning algorithms combined with traditional PID and Kalman filter-based systems ensure consistent and reliable data collection over extended periods, even under varying environmental conditions [4]. UAVs operating in challenging and remote locations for logistics, exploration, and data collection require stable sensors for reliable information gathering. Kalman filters, combined with machine learning, help maintain sensor stability, ensuring accurate and dependable data collection and contribute significantly in Delivery, Geological, and Meteorological Data Collection [9].

6. Conclusion

The stabilization of EO pods in UAVs is a complex but crucial aspect of ensuring high-quality imaging and precise targeting in various applications. The evolution of control techniques, from basic PID controllers to advanced methods like Kalman filters and machine learning, reflects the growing demands for stability and accuracy in increasingly dynamic and unpredictable environments. By integrating these advanced control algorithms, UAVs can maintain a stable LOS, even in challenging conditions, thereby enhancing their operational effectiveness across different sectors.

Traditional control methods such as PID controllers remain relevant due to their simplicity and effectiveness in many scenarios. However, their limitations in dealing with non-linearities and sudden disturbances necessitate the use of more advanced techniques. Kalman filters, with their ability to provide accurate state estimations from noisy data, and machine learning algorithms, which adapt to changing conditions through continuous learning, represent significant advancements in this field. The fusion of these techniques with traditional methods offers a balanced approach, combining the strengths of each to achieve optimal performance.

The application of these algorithms in scenario-specific contexts, such as military reconnaissance, target acquisition, and civilian aerial photography, demonstrates their versatility and critical role in enhancing UAV capabilities. As UAV technology continues to evolve, the development and refinement of stabilization algorithms will remain a key area of research, driving further improvements in UAV performance and expanding their range of applications.

By understanding and applying these advanced control strategies, engineers and researchers can continue to push the boundaries of what UAVs can achieve, ensuring that they remain at the forefront of technological innovation in both military and civilian domains.