TL;DR
This paper introduces a novel adaptive control strategy for a three-axis stabilized platform, enabling both precise low-speed tracking and wide-range scanning. By combining adaptive and nonsingular fast terminal sliding mode control, the system achieves faster convergence, reduced oscillations, and improved accuracy, even with unknown disturbances and input limitations.
What Problem Does This Paper Address?
Modern imaging systems, like those used in aerial surveillance and remote sensing, often require platforms that can perform diverse tasks. Specifically, a three-axis universal inertially stabilized platform (UISP) needs to excel at both high-precision, slow-speed video tracking and rapid, large-angle "swing scanning" for broader coverage. However, achieving this dual capability is challenging due to:
- Uncertainties: The platform's dynamics are affected by factors like friction, unmodeled dynamics, and external disturbances (e.g., from the aircraft it's mounted on), which are difficult to precisely model.
- Input Saturation: Actuators have physical limits on their output, meaning they can't respond instantaneously or infinitely to control commands.
- Conflicting Performance Needs: Low-speed tracking demands high accuracy and minimal jitter, while large-angle scanning requires fast, wide movements. Traditional control methods often struggle to balance these requirements.
Existing solutions like PID controllers are simple but lack robustness to uncertainties. Sliding mode control (SMC) offers robustness but can suffer from "chattering" (high-frequency oscillations) and requires knowledge of disturbance bounds. Terminal SMC can have singularity issues, and nonsingular variants often still need prior knowledge of disturbance bounds.
Methods
The paper proposes an Adaptive Nonsingular Fast Terminal Sliding Mode Control (ANFTSMC) strategy. Here's a breakdown of its key components:
- Dynamic Model: A second-order model of the UISP is established, incorporating friction, unmodeled dynamics, external disturbances, and importantly, input saturation. The non-smooth input saturation is approximated by a smooth hyperbolic tangent function for easier integration into the control design.
- Nonsingular Fast Terminal Sliding Mode Control (NFTSMC): This forms the core of the controller. It uses a specific sliding surface design that ensures the system's error converges to zero in a finite time, avoiding the singularity issues of earlier methods.
- Novel Reaching Law: To address chattering and improve convergence speed, a new reaching law is introduced. This law uses a modified normal distribution function to dynamically adjust the control gain based on the position and velocity errors. This allows for faster convergence when the system is far from the target and smoother control when it's close.
- Adaptive Control Law: To overcome the need for prior knowledge of disturbance bounds, an adaptive control law is designed. This law online estimates the parameters of the lumped uncertain disturbances, making the controller more robust and practical.
- Lyapunov Stability Theory: The stability of the entire ANFTSMC system is rigorously proven using Lyapunov theory, ensuring that the system remains stable and converges to the desired state.
Datasets
Not discussed in the paper. The study relies on simulations and experimental data from a physical three-axis UISP.
Key Insights and Ideas
- Bridging the Gap: The ANFTSMC effectively bridges the performance gap between precise low-speed tracking and dynamic large-angle scanning by intelligently adapting its control strategy.
- Disturbance Handling without Bounds: The adaptive component is crucial, allowing the controller to perform well even when the exact bounds of disturbances are unknown, a significant practical advantage.
- Smart Gain Adjustment: The novel reaching law is a clever way to manage the trade-off between fast convergence and chattering. By making the gain dependent on the error, it can be aggressive when needed and gentle when precision is paramount.
- Practicality through Approximation: Approximating the non-smooth input saturation with a smooth function simplifies the controller design and implementation without sacrificing significant performance.
Key Results
The paper presents extensive simulation and experimental results for a single-axis (A-gimbal) of the UISP.
- Improved Tracking Accuracy: The proposed ANFTSMC method consistently achieved smaller Mean Absolute Error (MAE) and Root Mean Square (RMS) values for both position and velocity tracking compared to PID and NFTSMC controllers in both sinusoidal (low-speed) and step (large-angle) signal tracking scenarios.
- Faster Convergence: For sinusoidal tracking, the ANFTSMC with a variable gain achieved convergence in 0.07s, significantly faster than the constant gain ANFTSMC (0.50s) and NFTSMC.
- Reduced Chattering: Simulations showed significantly less chattering with the proposed ANFTSMC compared to NFTSMC.
- Robustness: The method demonstrated good robustness to variations in system parameters (e.g., moment of inertia) and sudden external disturbances.
- No Overshoot: For step signal tracking, the ANFTSMC methods (both with variable and constant gain) exhibited no overshoot, a significant improvement over the PID controller.
Limitations and Future Work
- Single-Axis Validation: While the paper focuses on the A-gimbal, the UISP is a three-axis system. The authors mention that each axis can be considered independently, but a full three-axis experimental validation would be more comprehensive.
- Inherent System Delay: The authors note that the experimental results might be affected by inherent system delays, which were ignored for simplicity in some analyses.
- Attitude Compensation: Attitude compensation was not included in the simulations or experiments to ensure fairness in comparison, but it's a relevant aspect for real-world UISP applications.
Future work could involve extending the experimental validation to all three axes, incorporating attitude compensation, and potentially exploring further optimizations of the adaptive and reaching law components.
Summary updated