Graduates of the Mechatronics Engineering course at the Tecnológico de Monterrey, Santa Fe campus, have developed an algorithm that makes it possible to reduce the error rate in the relevant guidelines for mechatronics or robotics. This research was published in the journal “Expert Systems With Applications”, one of the world’s most prestigious scientific publications in the field of intelligent systems applied in industry, government processes and university activities.
The paper by Tec de Monterrey engineers Gerardo Ortiz, Daniel Blanck and Valentín Martínez, entitled “Optimal Neural Tuning of a Controller for a Parallel Robot”, explains that this finding becomes relevant as the research considers using open source, i.e. that any student or researcher interested in the topic can work with it, edit it, and even improve it.
“This paper deals with the gain adjustment of a PID controller for two brushless motors using three techniques: random assignment, DE and DNN. Three practical cases are analyzed to observe the advantages and disadvantages of each method. All measurements of position errors, velocity errors, and current are generated using the experimental prototype, saving the use of a mathematical model to solve the dynamics of the robot and actuators. Likewise, the effects of friction in the joints and closed kinematics,” the authors of the research specify in the conclusions of the published article.
IMPROVEMENT IN ACCURACY
Arguing the rationale behind their research, the Tec de Monterrey engineers explained some of the challenges they encounter in their daily work in industrial robotics.
In posing the problem, they pointed out that finding optimal controller gains for five-time parallel is a highly iterative process, that is, it requires repeating a set of actions multiple times with the intention of achieving a desired goal, objective, or result.
This iteration is necessary because many variables must be adjusted and often meet conflicting specifications such as energy efficiency and high accuracy. In addition, the nonlinear dynamic behavior must account for the motion limitations of the mechatronic or robotic system.
In addition, some drawbacks can be encountered when searching for optimal gains for a given controller, such as these optimal gains specific to the effector load, trajectory, and set of modeled parameters. In the presence of safety controls, changes in working conditions, tasks, parametric uncertainties and disturbances in the system, optimal settings lead to suboptimal settings, according to a report published in “Expert Systems With Applications.”
“The project originated in our industrial robotics class, where we were tasked with programming the motion trajectories of a robot to perform simple tasks such as transporting coins from one location to another. Subsequently, we corrected the processes of this differential evolutionary algorithm, achieving greater accuracy in the execution of instructions,” explained Gerardo Ortiz.
The software created with this effort can be easily interpreted by people with little knowledge in robotics and is characterized by its reproducibility and versatility for integration into different fields of knowledge, which is a contribution to national scientific research.
“In relation to artificial intelligence and machine learning, the value of our proposal is that it can be implemented in all technologies that involve an industrial process,” added Daniel Blanck.
The journal “Expert Systems With Applications” has a Journal Citation Reports (JCR) index of 8.5, which means it has an estimated average of at least 8 citations in the same year.
The article published in a scientific journal was developed in collaboration with professors: Héctor Cervantes, Enrique Chong and Carlos Cruz from the Mexico State campus.
Industrial robotics is the key to increasing productivity
Industrial robots are programmable devices that perform movements in three or more axes for various purposes. Industrial robotics is a branch of engineering that deals with the definition, industrial design, development and production of industrial robots that automate work. These robots perform their movements in a manufacturing or industrial production chain to perform tasks faster and more precisely, without rest and avoid long hours that endanger the health of workers. Its aim is nothing more than to increase productivity, use resources efficiently and reduce errors in the production chain.