In this study, a novel L-SHADE variant with collaborative scheme and comprehensive learning mechanism, named LSHADE-CLM, was proposed to improve the exploration and exploitation capabilities of the L-SHADE algorithm. In LSHADE-CLM, a novel cooperative mutation mechanism including "DEcurrent−to−pbetterr" and "DEcurrent−to−pbest−w1" is proposed in the mutation operation. In the "DEcurrent−to−pbetterr" strategy with comprehensive learning mechanism, the population covariance matrix is utilized to generate candidate solutions and guide the search direction. Meanwhile, a competitive reward mechanism is implemented to control the mutation factor F to generate a trial vector for the cooperative mechanism. Moreover, the dimensional reset strategy is applied to enhance the diversity of the population at the dimensional level when stagnation is identified at certain dimension. The proposed LSHADE-CLM is tested on the CEC2017 benchmark functions and compared with the other four state-of-the-art variants of L-SHADE. The experimental results demonstrated that the efficiency and effectiveness of the LSHADE-CLM algorithm for the non-separable optimization problem. © 2020 Elsevier B.V.