Computationally efficient hybrid differential evolution with learning for engineering application
Sanjoy Debnath, Wasim Arif, Debarati Sen, Srimanta Baishya · International Journal of Bio-Inspired Computation · 2022
Differential evolution (DE) is popular as an optimisation technique for its ability to achieve a globally optimal solution in a non-convex fitness landscape. However, convergence rate of DE is still unsuitable for real-time applications. Hence, a new leader-centric learning algorithm based on DE named hybrid differential evolution with learning (HDEL) is proposed in this work to improve the convergence performance of DE. In the proposed scheme, the mutation and crossover strategy is supervised by the learning knowledge of the global best and global worst individual as well as the personal best of an individual in its current generation. Extensive simulations have been performed on a set of 44 mathematical benchmark functions and a novel unmanned aerial vehicle (UAV) planning problem. The results of the performance comparison of the proposed HDEL with other state-of-the-art algorithms confirm the significant improvement of HDEL in terms of optimum value, convergence speed, and computational complexity.