Work

Student Psychology-Based Optimization Algorithm Enhanced by Teaching–Learning Mechanism for Global Optimization and Real-World Issues

Year: 2026

Type: article

Abstract: Student Psychology-Based Optimization (SPBO) provides a simple human-inspired search framework, but its dependence on random subject-wise learning behaviors, limited peer interaction, and insufficient utilization of historical information may restrict its performance on complex multimodal optimization problems. To address these limitations, this study proposes a Multi-Strategy Improved Student Psychology-Based Optimization algorithm, termed MLSPBO. Three complementary learning mechanisms are incorporated into the original framework. First, a multi-teacher hierarchical guidance strategy constructs a rank-weighted collective teaching reference to reduce excessive dependence on a single elite individual and provide diversified search directions. Second, a rank-adaptive peer mutual-learning strategy enables individuals with different performance levels to adopt differentiated learning behaviors, thereby enhancing information exchange and population diversity. Third, a self-reflection and remedial learning strategy exploits historical best information and search stagnation to strengthen local refinement and escape from unfavorable regions. The performance of MLSPBO is evaluated on the CEC2020 and CEC2022 benchmark suites under 10- and 20-dimensional settings and compared with nine representative metaheuristic algorithms. MLSPBO obtains the best or tied-best mean results on 7/10 and 9/10 CEC2020 functions and on 11/12 functions under both CEC2022 settings. Its Friedman mean ranks are 1.30, 1.10, 1.33, and 1.17, respectively, consistently ranking first among all compared methods. In addition, the Wilcoxon signed-rank test yields an aggregated 39 wins, 3 ties, and 2 losses against the original SPBO. The practical performance of MLSPBO was further evaluated using a three-dimensional robotic arm path planning problem. Its mean cost was 291.34, approximately 5.44% lower than that obtained by NDO and 6.10% lower than that of the original SPBO. The results suggest that the proposed method is not limited to numerical benchmark optimization and can also provide reliable solutions for constrained robotic path-planning tasks.

Source: Mathematics

DOI: 10.3390/math14203683

Authors: Rutong Li, Kou Lei

Institutions: University College London, Shandong University

Language: English

Topic: Metaheuristic Optimization Algorithms Research

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

SDG: Quality education

Open Access status: gold

Estimated APC paid: $3,137