TL;DR
This paper presents a novel flight control system for unmanned aerial vehicles (UAVs) to loiter over a ground moving target (GMT) whose maneuvers are unpredictable. The system combines a discrete-time Lyapunov guidance law with an integral sliding mode controller and a Rao-Blackwellised particle filter to simultaneously estimate the GMT's motion and maneuver, enabling robust loitering even with unknown target behavior.
What Problem Does This Paper Address?
Tracking and loitering over a ground moving target (GMT) is a critical capability for various applications, including surveillance and border patrol. A key challenge arises when the GMT's movement is unpredictable or "unknown," meaning its future maneuvers cannot be precisely foreseen. This unpredictability makes it difficult for a UAV to maintain a stable loitering position at a desired distance. Traditional approaches often assume known target dynamics or rely on continuous-time models, which are not directly applicable to discrete-time systems with unknown, stochastically varying maneuvers. This paper aims to bridge this gap by developing a robust discrete-time control strategy that can handle such uncertainties.
Methods
The proposed solution is a three-pronged approach:
- Discrete-Time Lyapunov Guidance Vector Field: This component generates the desired trajectory for the UAV. Unlike previous methods that focused on continuous-time systems, this work extends the Lyapunov guidance vector field concept to a discrete-time setting, ensuring asymptotic convergence to a circular orbit around the GMT.
- Discrete-Time Integral Sliding Mode Control (ISMC): To ensure the UAV accurately tracks the guidance commands despite disturbances and uncertainties, an ISMC is employed. This controller is designed to be robust and handles the discrete-time nature of the system.
- Rao-Blackwellised Particle Filter (RBPF): This is a crucial component for estimating the GMT's state. The RBPF is used to simultaneously estimate both the GMT's motion (position and velocity) and its maneuver mode. The maneuver is modeled as a finite-state Markov chain. This approach is more efficient than standard particle filters, requiring fewer particles for accurate estimation. The system leverages the principle of certainty equivalence, using the estimated GMT state to drive the guidance and control laws.
Datasets
Not discussed in the paper. The evaluation is performed through simulations.
Key Insights and Ideas
- Discrete-Time Guidance is Tricky: Adapting continuous-time guidance laws to discrete-time systems introduces complexities, particularly regarding sampling frequency requirements to ensure effectiveness.
- RBPF for Unknown Maneuvers: The Rao-Blackwellised Particle Filter is a powerful tool for simultaneously estimating both the state and the hidden Markov state (maneuver mode) of a target, offering a significant advantage when target dynamics are unknown.
- Certainty Equivalence for Robustness: Using estimated states in control laws (certainty equivalence) is a practical approach when true states are unavailable, and the paper demonstrates its effectiveness in this context.
- ISMC for Disturbance Rejection: Integral Sliding Mode Control provides strong robustness against system uncertainties and external disturbances, making it suitable for real-world scenarios where perfect models are unattainable.
Key Results
- Simulations demonstrate that the proposed ISMC controller achieves the shortest setting time and zero steady-state error in tracking guidance commands when the GMT's state is known, outperforming PID and standard SMC.
- The RBPF shows superior performance in estimating the GMT's state compared to an Extended Kalman Filter (EKF), even when the transition probabilities of the GMT's maneuver are unknown. This holds true for both camera and radar sensor models.
- When using estimated GMT states, the proposed controller successfully enables the UAV to loiter around a stationary GMT with the desired radius.
- The system is shown to be effective in a more complex 6-DOF UAV model, demonstrating its potential for real-world applications.
Limitations and Future Work
- Certainty Equivalence Stability Proof: The authors acknowledge that proving the stability of the overall system when using estimated states from a nonlinear filter (like the RBPF) is challenging due to the difficulty in computing exact error bounds. While simulations are promising, rigorous theoretical guarantees for the estimated-state scenario are not provided.
- Computational Resources: While the RBPF is more efficient than standard PFs, the computational demands of particle filters can still be a consideration for real-time implementation on resource-constrained UAVs.
- Future Work: The paper focuses on a fixed-wing UAV and a simplified maneuver model. Future work could explore more complex UAV dynamics (e.g., multirotor), more sophisticated maneuver models, and potentially extend the approach to cooperative multi-UAV scenarios.
Summary updated