Performance Evaluation of New Hybrid Variant of Teaching Learning Based Optimization Algorithm for Design of Muti-objective Eco-driving Train Speed Trajectory

The no-free-lunch theorem for optimization indicates that the effectiveness of an optimization method is heavily influenced by the specific problem it aims to solve. The Teaching Learning Based Optimization (TLBO) algorithm is recognized as one of the most straightforward metaheuristic optimization methods since it does not require fine-tuning of specific parameters. However, its effectiveness in optimizing train speeds remains underexplored. This paper assesses the efficacy of four TLBO variants: the original TLBO, TLSBO, BTLBO, and a novel hybrid variant that integrates differential evolution (DE) with the original TLBO, named sDTLBO. This study tackles the multi-objective challenge of establishing an eco-driving speed trajectory for a train, emphasizing energy consumption, punctuality, and jerkiness. In the proposed sDTLBO, the teacher phase of TLBO is enhanced by the mutation operator of DE method. We compare the performance of the four TLBO variants against four other metaheuristic algorithms: Particle Swarm Optimization (PSO), Genetic Algorithms (GA), DE, and Artificial Bee Colony (ABC), via numerical simulations. The simulation results demonstrate that the novel sDTLBO outperforms all tested variants in all key aspects: energy consumption efficiency, punctuality, and jerkiness, across various maximum iterations used. Thus, sDTLBO emerges as a potent variant for exploring and optimizing eco-driving speed for rail vehicles.