Abstract
The fixed-time control for systems with time-varying parameters is investigated in this article. Different from the previous pure robust control, the proposed method is less conservative. Applying this method could avoid the existence of unknown time-varying parameter derivatives in control design. Combining with adaptive fuzzy control and projection operator, a novel singularity-avoidance virtual controller is constructed via backstepping. The proposed controller guarantees that the tracking errors converge to a user-defined precision within a prescribed fixed-time. Two examples are used to show that this method can achieve fixed-time control even if the time-varying parameters are unknown and non-asymptotically constant.
Introduction
In robotics and aerospace engineering, mechanical systems rarely possess constant parameters. Factors such as varying payloads on a robot arm or temperature-dependent friction create "time-varying" environments. Traditional adaptive control strategies typically assume these parameters are constant or change very slowly. However, when parameters fluctuate rapidly and persistently, standard controllers often fail to achieve high-precision tracking.
This research addresses the challenge of fixed-time stability. Unlike asymptotic stability (where error eventually reaches zero) or finite-time stability (where convergence time depends on initial conditions), fixed-time stability ensures the system converges within a maximum time limit ($T_{max}$), regardless of the initial state. Achieving this for time-varying systems is difficult because it usually requires knowledge of the parameter's rate of change (derivative), which is rarely available in practice.
Methods
The proposed approach utilizes a multi-layered strategy to achieve fixed-time convergence without requiring parameter derivatives.
Nominal Substitution (Congelation of Variables)
Instead of attempting to track a continuously moving parameter $\theta(t)$, the researchers replace it with a constant nominal value ($l_\Theta$). The difference between the actual parameter and this nominal guess is treated as a bounded disturbance. This "congelation" eliminates the need to calculate unknown derivatives.
Adaptive Fuzzy Logic Systems (FLS)
To compensate for the "interference term" created by substituting a constant for a variable, Type-2 Fuzzy Logic Systems are employed. These systems approximate and cancel out complex, unknown disturbances that arise from the parameter fluctuations.
Singularity-Avoidance and Backstepping
Fixed-time control often suffers from "singularities"—mathematical points where the control law might require division by zero as the system nears its target. The authors introduced a novel switched function ($sg_i$) and a modified quadratic Lyapunov function to ensure smooth control. This was integrated into a recursive backstepping design, using a projection operator to keep adaptive estimates within physically realistic bounds.
Results and Discussion
The methodology was validated through numerical simulations and a physical pendulum model where gravity ($g$) and friction ($K$) were oscillated to mimic instability.
Key Insights
- The Nominal Shortcut: Treating the gap between real parameters and nominal guesses as a bounded disturbance allows intelligent fuzzy logic to handle the complexity without needing derivative data.
- Independence from Initial States: The fixed-time approach ensures the system settles within a user-defined window (e.g., 0.65 seconds) regardless of the starting position.
- User-Defined Precision: Engineers can set a specific precision interval ($[-\pi, \pi]$); the system is guaranteed to enter and remain within this window in fixed time.
Performance
In simulations, the controller achieved a settling time of approximately 0.65 seconds. Compared to traditional robust methods, the proposed controller produced significantly smoother control inputs with less "chatter," which is critical for reducing mechanical wear on actuators.
Conclusion
This paper demonstrates a less conservative, high-precision method for controlling nonlinear systems with unknown, non-asymptotically constant parameters. While the mathematical design is intensive, it provides a robust framework for fixed-time convergence. Future work will focus on reducing the complexity of the controller for implementation on lower-power hardware and incorporating internal model knowledge to further enhance efficiency.
Summary updated