In this paper, the synthesis and the properties of the neural speed controller trained online are presented. The structure of the controller and the training algorithm are described. The resilient backpropagation (RPROP) algorithm was chosen for the training process of the artificial neural network (ANN). The algorithm was modified in order to improve controller operation. The specific properties of the controller, i.e., adaptation and auto-tuning, are illustrated by the results of both simulation and experimental research. An electric drive with permanent magnet synchronous motor (PMSM) was chosen for experimental research, due to its impressive dynamics. The obtained results indicate that the presented controller may be implemented in industrial applications.

This paper is dedicated to the motion control system of a multimass mechanism with occurring resonances and variable moment of inertia. The proposed concept of control structure for such mechanism consists of a filter damping higher resonance frequencies and an adaptive speed controller. The filter has properly selected fixed characteristics giving a compromised filtering effect in a range of resonance frequency variation, caused by the variable moment of inertia. The used neural adaptive controller realizes active damping of lower resonance frequencies, which are not eliminated by the resonance filter. This original concept is compared with a more complex solution, treated as a reference, which contains an offline-tuned resonance filter together with an offline-adjusted PI-two degrees of freedom controller. Good speed control dynamic properties of the proposed concept in comparison with the reference complex solution are proven by laboratory results.

This paper deals with the problem of robust speed control of electrical servodrives. A robust speed controller is developed using an artificial neural network (ANN), which creates a nonlinear characteristic of controller. An original method of neural controller synthesis is presented. The synthesis procedure is performed in two stages. The first stage consists in training the ANN and at the second stage controller settings are adjusted. The use of the proposed controller synthesis procedure ensures robust speed control against the variations of moment of inertia and stator magnetic flux. Simulations and laboratory results validate the robustness of the servodrive with permanent magnet synchronous motor

Purpose: The aim of the research was to find out a method of adaptive speed control robust against variation of selected parameters of system like moment of inertia, time constant of torque control loop or torque coefficient of the motor.Design/methodology/approach: The main goal of the research was achieved due to application of artificial neural network (ANN), which was trained on line on the base of speed control error. The good results were gained by elaboration of enough fast and precise training algorithm and proper ANN structure.Findings: The work shows a structure of artificial neural network (ANN), applied as adaptive speed controller, and presents an algorithm of ANN training. Some versions of this algorithm were analysed and verified by simulation and experimental tests.Research limitations/implications: The research should be continued to determine a final version of training algorithm and its influence on controller properties.Practical implications: The elaborated adaptive controller can be easily used by applying microprocessor system available now on the market. The proposed control solution is robust against parameters variation as well as their imprecise identification. The controller has ability of self‐tuning which can have great practical advantage.Social implications: Social implications are difficult to determine.Originality/value: The paper presents a new solution of adaptive speed controller, which means a new ANN structure and new training algorithm.

This paper presents an application of an Unscented- and a Fuzzy Unscented- Kalman Filter (UKF and FUKF) to the estimation of mechanical state variables and parameters in a drive system with an elastic connection. The cascade control structure incorporating an IP controller supported by two additional feedbacks and suitable adaptation mechanism is investigated in this study. The coefficients of the control structure are retuned on the basis of the value of mechanical parameters estimated by filter. The effectiveness of the proposed approaches (classical and fuzzy) is researched through simulation and experimental tests.

The monitoring of presence is a timely topic in intelligent building management systems. Nowadays, most rooms, halls, and auditoriums use a simple binary presence detector that is used to control the operation of HVAC systems. This strategy is not optimal and leads to significant amounts of energy being wasted due to inadequate control of the system. Therefore, knowing the exact person count facilitates better adjustment to current needs and cost reduction. The vision-based people-counting is a well-known area of computer vision research. In addition, with rapid development in the artificial intelligence and IoT sectors, power-limited and resource-constrained devices like single-board computers or microcontrollers are able to run even such sophisticated algorithms as neural networks. This capability not only ensures the tiny size and power effectiveness of the device but also, by definition, preserves privacy by limiting or completely eliminating the transfer of data to the cloud. In this paper, we describe the method for efficient occupancy estimation based on low-resolution thermal images. This approach uses a U-Net-like convolutional neural network that is capable of estimating the number of people in the sensor’s field of view. Although the architecture was optimized and quantized to fit the limited microcontroller’s memory, the metrics obtained by the algorithm outperform the other state-of-the-art solutions. Additionally, the algorithm was deployed on a range of embedded devices to perform a set of benchmarks. The tests carried out on embedded processors allowed the comparison of a wide range of chips and proved that people counting can be efficiently executed on resource-limited hardware while maintaining low power consumption.