Bidirectional teaching and peer-learning particle swarm optimization

• A PSO algorithm’s variant, abbreviated as BTPLPSO, is proposed.
• BTPLPSO aims to more accurately model the real-world teaching–learning scenario.
• Enhanced teaching and peer-learning framework is used to improve PSO’s performance.
• OEDELS module is proposed as efficient learning strategy for global best particle.
• BTPLPSO outperforms its peers in term of searching accuracy and convergence speed.

Most of the well-established particle swarm optimization (PSO) variants do not provide alternative learning strategies when particles fail to improve their fitness during the searching process. To solve this issue, we improved the state-of-art teaching–learning-based optimization algorithm and adapted the enhanced framework into the PSO. Thus, we developed a bidirectional teaching and peer-learning PSO (BTPLPSO). Specifically, the BTPLPSO uses two learning phases, namely, the teaching and peer-learning phases. The particles first enter the teaching phase and update their velocity based on their personal and global best information. However, when particles fail to improve their fitness in the teaching phase, they enter the peer-learning phase and learn from the selected exemplar. To establish a two-way learning mechanism between the global best particle and the population, we developed an orthogonal experimental design-based elitist learning strategy to improve the global best particle by fully exploiting the useful information of each particle. The proposed BTPLPSO was thoroughly evaluated on 25 benchmark functions with different characteristics. The simulation results confirmed that BTPLPSO significantly outperforms eight well-established PSO variants and six cutting-edge metaheuristic search algorithms.