A Hybrid Bio-Inspired Metaheuristic Applied to PID Controller Parameter Tuning

Leonardo Ramos Rodrigues, João P. P. Gomes2022 · DOI

TL;DR

This paper introduces CSASOS, a novel hybrid bio-inspired metaheuristic that combines the exploration strengths of the Crow Search Algorithm (CSA) with the exploitation strengths of the Symbiotic Organisms Search (SOS). CSASOS is applied to the challenging problem of tuning adaptive PID controllers, demonstrating superior performance compared to its individual components in achieving optimal control parameters.

What Problem Does This Paper Address?

Tuning Proportional-Integral-Derivative (PID) controllers is a critical task in many industrial applications. While standard PID controllers are widely used, their performance can be significantly enhanced by employing adaptive gains, particularly those based on smooth Gaussian functions. However, designing these adaptive PID controllers involves optimizing a complex set of parameters, which can be a computationally intensive task. Traditional optimization methods may struggle with the non-linearity and complexity of this problem. This paper addresses the challenge of efficiently and effectively tuning these adaptive PID controllers by proposing a new hybrid metaheuristic approach.

Methods

The core of this paper is the introduction of the CSASOS (Crow Search Algorithm - Symbiotic Organisms Search) algorithm. CSASOS is a multi-population metaheuristic designed to leverage the complementary strengths of two bio-inspired algorithms:

  • Crow Search Algorithm (CSA): Known for its ability to explore the search space and identify promising regions. In CSASOS, one subpopulation uses CSA's search rules to broadly scan for potential solutions.
  • Symbiotic Organisms Search (SOS): Recognized for its effectiveness in exploiting promising regions and refining solutions. The second subpopulation in CSASOS employs SOS's search rules for a more focused and precise optimization.

The hybridization strategy involves dividing the search agents into two subpopulations. The CSA subpopulation is initially larger to promote exploration. Periodically, the best feasible solutions found by CSA are transferred to the SOS subpopulation, which then performs a refined search. Notably, CSASOS utilizes only the mutualism and commensalism phases of SOS, as the parasitism phase (which is more for diversification) is handled by the CSA. Opposition-based learning is used for initializing the CSA subpopulation to improve initial diversification. For handling bound constraints, the "rand base" method is employed.

Datasets

Not discussed in the paper. The evaluation is performed using a simulated DC motor system.

Key Insights and Ideas

  • Synergy of Exploration and Exploitation: The paper highlights how combining a strong exploration algorithm (CSA) with a strong exploitation algorithm (SOS) can lead to a more robust and efficient optimization process than using either algorithm alone.
  • Multi-Population Advantage: The multi-population approach allows for parallel exploration and exploitation, with a controlled transfer of promising solutions between subpopulations to guide the search effectively.
  • Adaptive PID Tuning is Complex: The problem of tuning adaptive PID controllers, with its multiple parameters and performance metrics (like ITAE), is a suitable testbed for advanced metaheuristics.
  • CSASOS Outperforms Components: The results strongly suggest that the hybrid nature of CSASOS provides a significant advantage in finding better solutions and achieving more consistent results compared to standalone CSA or SOS.

Key Results

The CSASOS algorithm was tested on an adaptive PID controller parameter tuning problem for a DC motor speed control loop. The performance was evaluated using the Integral Time Absolute Error (ITAE) metric.

  • Superior Solution Quality: CSASOS consistently achieved the lowest average ITAE value (3.71 ± 0.14 × 10⁻⁴) compared to CSA (4.57 ± 0.41 × 10⁻⁴) and SOS (3.97 ± 0.22 × 10⁻⁴).
  • Improved Consistency: CSASOS also demonstrated the smallest standard deviation in ITAE values, indicating that it reliably finds good solutions across multiple runs.
  • Adaptive PID Advantage: The paper also shows that the adaptive PID controller designed using CSASOS provides a significantly better step response (lower overshoot and peak time) compared to a standard PID controller, even when the standard PID meets basic design requirements.

Limitations and Future Work

  • Scope of Application: The authors suggest that future research could extend the application of CSASOS to other complex optimization problems, including multi-objective and binary optimization problems.
  • Communication Strategy: Investigating new strategies for guiding the communication and information exchange between the two subpopulations in CSASOS is identified as a promising area for future work.
  • Hyperparameter Tuning: While experimental tests were used to define hyperparameters like migration period (MP) and migration size (MS), further systematic studies on their impact could be beneficial.

Summary updated

BibTeX
@misc{rodrigues2022hybrid,
  author = {Leonardo Ramos Rodrigues and João P. P. Gomes},
  title = {{A Hybrid Bio-Inspired Metaheuristic Applied to PID Controller Parameter Tuning}},
  year = {2022},
  doi = {10.46254/eu05.20220295},
}
Figure