Abstract
This paper discusses a Takagi-Sugeno (T-S) fuzzy controller design problem for a phugoid model-based multi-agent system. The error between the state of a phugoid model and a reference is defined to construct a multi-agent system model. A T-S fuzzy model of the multi-agent system is built by introducing a nonlinear controller. A fuzzy controller is then designed to stabilize the T-S fuzzy model, where the synthesis condition is represented in terms of linear matrix inequalities.
Introduction
Coordinating a "swarm" or multi-agent system (MAS) is a significant challenge in robotics, especially when the agents follow complex physical laws. Most existing MAS research uses simplified "point-mass" models. However, real-world vehicles like underwater gliders or UAVs exhibit phugoid motion—long-period oscillations where potential energy (altitude) and kinetic energy (speed) are exchanged. Controlling these in a formation is difficult due to inherent nonlinearities, scalability issues for centralized controllers, and the coordination requirements for maintaining specific relative distances and velocities without a global supervisor.
Methodology
The researchers propose a multi-step control architecture to address these challenges. First, they define "tracking error" states—the difference in position and velocity between an agent $k$ and its predecessor $k-1$. A preliminary nonlinear controller is introduced to cancel out complex terms in the flight equations, simplifying the system's structure.
The remaining nonlinear system is then represented using Takagi-Sugeno (T-S) fuzzy modeling via the "sector nonlinearity" method. This represents the system as a weighted combination of linear subsystems, allowing the application of linear control theory. Finally, stability conditions are expressed as Linear Matrix Inequalities (LMIs), which can be solved numerically to guarantee controller stability. The approach was validated through a numerical simulation of three agents with parameters ($L=300, m=24kg$) representative of medium-sized autonomous gliders.
Results and Discussion
The system utilizes a decentralized "follow-the-leader" logic, where each agent only requires information from the agent immediately preceding it. This enhances robustness against single-point failures. The mathematical proof (Theorem 1) demonstrates that if the LMI conditions are met, the formation error converges to zero.
In simulations, the three agents successfully converged to their assigned states. As shown in the state plots, the variables $x_{1,k}$ through $x_{5,k}$ stabilize over a 10-second interval, confirming the effectiveness of the T-S fuzzy control strategy in bridging the gap between high-level formation theory and low-level aircraft longitudinal dynamics.
Conclusion
While successful in 2D longitudinal motion, future work is needed to extend this model to 3D maneuvers involving rolling and yawing. Additionally, the current model assumes equilibrium between thrust and drag, which may not hold in turbulent conditions. Future research should also address the impact of noisy communication or time delays on formation stability.
Summary updated