Benchmark of PID Neural Speed Controllers for Permanent Magnet DC Motors on an Artificial Intelligence Embedded System

In industrial automation, the PID controller plays a crucial role as it allows for process control, making it a widely utilized tool in automation. However, due to technological advancements and the continuous growth of the industry, simple PID controllers with straightforward architectures are becoming obsolete. This has led to the current situation where efforts are being made to find new, more robust, and optimal PID control methods, which has resulted in the development of new techniques based on neural networks and bio-inspired algorithms. These techniques offer advantages in terms of production line efficiency, but selecting the right controller for nonlinear industrial systems like DC motors can be challenging. To address this challenge, an alternative is presented involving two PID controller architectures based on neural networks. In one, the neural network adjusts the PID controller parameters, while in the other, the neural network and PID controller collaborate to regulate the speed of a DC motor in a training environment. These controllers are developed using tools such as Python, Linux, VSCode, and Arduino, along with the ARM Jetson Xavier NX platform and its Jet pack 5 development environment. Performance comparison will rely on performance indices like the Integral of Absolute Error (IAE), supported by statistical analysis using Microsoft Excel. The results hold promise for the field of industrial control.