The title is concise and describes the article clearly, but it should specify that this method was tested in linear time-invariant systems.

Abstract The problems addressed by the paper are clearly defined, in this case, multi-objective optimization. The method used by the authors to tune the controller enlarges the readers´ knowledge about the capability of classic PID control.

Although some references are not recent, they support the goal of the research. There are references about the advances of the PSO algorithm, criteria to design PID controllers, and also, is cited how these two methods have been integrated.

The introduction describes the problem of tuning PID controllers and how this problem has been solved with traditional control engineering tools. Then, the authors cite how evolutionary computing algorithms have been used to improve the PID controller's performance. This improvement is made by addressing the tuning problem as a many-objective optimization problem seeking for optimal overall performance.

The structure of the methodology is clear, and even a reader with limited or no knowledge about the subject can follow and understand how the research was conducted. First, the authors give a brief background of PID control and Particle Swarm Optimization; then, the objectives to evaluate the plant performance are correctly defined and include both measurements of control engineering and measurements of PSO.

The problem statement is correctly developed. The five objectives are consistent with the metrics described in the previous sections.

The use of 4 different types of plants to prove the generality of the proposed method is good practice. These plants include time delay and double lag (Gp1), a 4-lag plant (Gp2), a fourth-order system (Gp3), and a non-minimum phase plant (Gp4).

However, it is not clear how objectives 1, 2, and 3 were evaluated. It seems that they were run in independent experiments, which is inconsistent. I think that way because it is written that "a step is applied solely to..." in the first three objectives. This issue complicates the replication of the study. Furthermore, the procedure to avoid unstable tunings, mentioned in section 3, is poorly described.

The results are presented clearly through the use of graphs, but the interpretation could be improved. The results shown in table 2 are not explained clearly. The solution space with normalized results is shown in Figure 7. If the first four objectives are evaluated according to the integral of absolute error (which we are trying to minimize), how do we interpret a normalized value of 1? Is it a good or bad value?

In table 1, two PID solutions values for Gp1 are shown (Solutions 17 and 18). Solutions 17 and 18 obtained a similar value for objective 4; however, the values obtained for objectives 1, 2, and 3 are very different, which is inconsistent. Proper performance respect to Set-point tracking, Load disturbance rejection, and Output disturbance rejection, generally is associated with high control effort. So, how was obtained a very different behavior with a similar control signal? These contradictions have to be explained clearly.

Conclusions do not emphasize the flexibility of the proposed method to achieve different objectives simultaneously over the classic tuning methods. Nor is it mentioned in this section that this method is limited to SISO systems. Finally, future work should be directed to non-linear systems or MIMO systems.

Overall strengths - This method allows the control engineer to choose the parameters Kp, Ki, and Kd according to the process requirements. - The proposed objectives improved the PID control performance by including the actuation cost.

Specific weaknesses 1. The procedure to avoid unstable tunings, mentioned in section 3, is poorly described. 2. Nor is it mentioned how the maximum and minimum values for the parameters Kp, Ki, and Kd were established in the PSO algorithm.