Automated processes require groups of controllers to manipulate the process variables, in order to reduce the margin of error between an established reference value and the value marked by the measurement instruments.One of these controllers is the PID (Proportional, Integral and Derivative) which reduces the error from a feedback given by a sensor.This control strategy depends on three constants, whose values are estimated in different ways depending on the particular characteristics of each process.Sometimes, this feature limits the effectiveness of the controller, since it depends on a mathematical approach and a simulation.An alternative solution to this limitation is proposed in this paper, which allows to estimate the constants of a PID controller automatically without needing to know the process or to perform a previous simulation.The value of the constants was estimated using a bioinspired optimization algorithm, which is fed back from the process information by means of a data acquisition card.Said card is a physical controller that has the function of regulating an actuator from the information of a sensor.Keyword-Evolutionary algorithm, PID, controller, process variable, feedback. I. INTRODUCTIONThe feedback controllers have been widely studied in the area of automation and control, since they perform various operations to reduce the margin of error between an expected value and one measured by a sensor in an automated process [1].There are several techniques to reduce the margin of error, ranging from the design of controller architectures to approximate mathematical models of the processes to be controlled.However, the effectiveness of these techniques decreases depending on the configuration of the controller, since, when trying to incorporate a controller in different processes, its configuration must be changed [2][3].Despite this, techniques based on the designer's expertise or a simulation are still implemented, since at an industrial or academic level they are flexible enough to adapt to various types of processes.However, when implementing these techniques, environmental effects are not considered in the process, which causes the controller to be periodically readjusted.This limitation has been partially solved with industrial-type controllers, which have a self-configuration function [2][3].A disadvantage of implementing this type of controllers is that they have a high cost and increase the number of components that must be used to control a process.This characteristic has allowed exploring other areas to contribute in the search for a solution to this problem, among which is machine learning (4).Machine learning is studied in different branches of engineering, thanks to its versatility to be implemented in the solution of various types of problems.Normally supervised and unsupervised learning techniques are studied, which try to emulate the characteristics of human beings to reason and solve problems [5].These techniques have been very popular in industrial applications, for example: the systems of classification of raw material by means of artificial vision, the prediction of faults by digital image processing systems or automated irrigation systems that are reconfigured depending on the season of the year [6][7][8].Evolutionary algorithms have become one of the studied subjects of supervised learning, which emulate the behavior of Darwinian evolution through different computational algorithms.It can be said that this type of technique is inspired by natural evolution and allows finding several solutions to a problem from a set of initial solutions.This set undergoes modifications when applying to each individual an operator of selection, recombination or mutation, by means of which its aptitude value is improved so that it becomes the most suitable [9,10].Consequently, this paper describes a partial solution to the feedback controller limitation by means of a bioinspired optimization strategy, which allows to automatically configure a PID-type controller.The advantage of this strategy is that it does not require prior knowledge of the system to be controlled, therefore, it is flexible enough to generalize to any type of programmable controller.The characteristics described above are organized