Introduction
In control engineering, time delays are often referred to as "stability killers." Whether originating from sensor lag or actuator response times, these delays can cause nonlinear systems—where outputs are not directly proportional to inputs—to oscillate or fail. Managing these time-varying delays is particularly challenging in complex systems like a truck-trailer backing up.
To address this, researchers often employ Takagi-Sugeno (T-S) fuzzy models to approximate nonlinear systems through local linear rules. However, traditional "delay partitioning" methods, which divide the delay period into smaller segments for analysis, frequently suffer from high computational complexity or overly conservative results that limit controller efficiency.
The Exponentially Narrow Delay Interval Technique
This study introduces a novel "exponentially narrow delay interval" technique designed to partition time-varying delays more effectively. Unlike standard methods that divide intervals into equal parts, this approach uses a recursive formula governed by a single tunable parameter, $p \in [0, 1]$.
By adjusting $p$, researchers can "zoom in" on specific segments of the delay range $[d_0, d_2]$. This flexibility allows for a more precise characterization of the delay's behavior, making the optimization process significantly more computationally efficient than traditional multi-parameter partitioning schemes.
Mathematical Innovations: Parameter-type Reciprocally Convex Inequality
Beyond interval partitioning, the authors propose a new "parameter-type reciprocally convex inequality" to estimate the derivative of the Lyapunov-Krasovskii functional (LKF). Stability analysis typically relies on LKFs, but estimating their derivatives often involves mathematical "gaps" that lead to less precise bounds.
The introduced inequality utilizes pre-specified parameters ($v_1, v_2$) to provide tighter bounds. Crucially, it eliminates certain nonlinear constraints and "slack matrices" that typically increase the number of decision variables, thereby simplifying the resulting Linear Matrix Inequalities (LMIs) without sacrificing accuracy.
Application to the Truck-Trailer System
The proposed strategies were validated using a benchmark truck-trailer system, a three-state nonlinear model representing the angular difference between the truck and trailer, the trailer's angle, and its vertical position. The primary metric for success was the Maximum Admissible Upper Bound (MAUB)—the longest delay the system can withstand while remaining stable.
Key Results
- Enhanced Delay Tolerance: With the tuning parameter set to $p=0.4$, the system achieved a MAUB of 276.7101. This represents a significant improvement over previous state-of-the-art results (e.g., 262.4908 from Lian et al., 2021).
- Robust Stability: Simulations confirmed that the truck-trailer system successfully converged to a stable state even under oscillating, time-varying delays ($d(t) = 8 + 0.2 \sin(t)$), where older methods might have reached instability.
Conclusion and Future Work
The exponentially narrow delay interval technique provides a powerful "tuning knob" for control engineers, allowing them to utilize more information about delay behavior in real-time. While the parameter $p$ simplifies the optimization process, future research could focus on automating the selection of the optimal $p$ for specific systems. Additionally, moving from numerical simulations to physical hardware testing will be essential to evaluate the technique's performance against real-world noise and jitter.
Summary updated