Monitoring and classification of dairy cattle behaviours is essential for optimising milk yields. Early detection of illness, days before the critical conditions occur, together with automatic detection of the onset of oestrus cycles is crucial for obviating prolonged cattle treatments and improving the pregnancy rates. Accelerometer-based sensor systems are becoming increasingly popular, as they are automatically providing information about key cattle behaviours such as the level of restlessness and the time spent ruminating and eating, proxy measurements that indicate the onset of heat events and overall welfare, at an individual animal level. This paper reports on an approach to the development of algorithms that classify key cattle states based on a systematic dimensionality reduction process through two feature selection techniques. These are based on Mutual Information and Backward Feature Elimination and applied on knowledge-specific and generic time-series extracted from raw accelerometer data. The extracted features are then used to train classification models based on a Hidden Markov Model, Linear Discriminant Analysis and Partial Least Squares Discriminant Analysis. The proposed feature engineering methodology permits model deployment within the computing and memory restrictions imposed by operational settings. The models were based on measurement data from 18 steers, each animal equipped with an accelerometer-based neck-mounted collar and muzzle-mounted halter, the latter providing the truthing data. A total of 42 time-series features were initially extracted and the trade-off between model performance, computational complexity and memory footprint was explored. Results show that the classification model that best balances performance and computation complexity is based on Linear Discriminant Analysis using features selected through Backward Feature Elimination. The final model requires 1.83 ± 1.00 ms to perform feature extraction with 0.05 ± 0.01 ms for inference with an overall balanced accuracy of 0.83.

There are various tools for real-time data acquisition and analysis of high-frequency signals. Usually, the tools for data acquisition and analysis are incompatible with each other, expensive, or require the utilization of various technologies. Designing a unique toolset for real-time data acquisition and analysis of high-frequency signals presents a large challenge, both in terms of hardware and software design. Some of the main challenges include fulfilling strict real-time performance requirements, using cost-effective hardware components, and designing a software architecture with minimal software latency for data sampling on an embedded platform and data analysis on a PC machine. This paper presents a software architecture design methodology for embedded platforms, including an RTOS and an Ethernet interface, along with the application architecture on the host side. To demonstrate the advantages of the proposed methodology, a prototype consisting of the STM32H747 microcontroller, FreeRTOS operating system, and a GUI application based on the Qt framework has been created. It has been shown that using the proposed architecture design methodology makes it possible to preserve the maximal hardware sampling rate utilizing a 100 Mbps Ethernet link. In the experiments, sampling rates up to 4.57 MSps for 10-bit resolution and real-time transfer of data over Ethernet link and visualization on host PC have been achieved with this specific MCU.

Recent studies have focused on accuracy as the key state of charge (SoC) estimation algorithms’ performance metrics, whereas just a few of them compare algorithms in terms of energy efficiency. Such a comparison is important when selecting an algorithm that should be implemented on a resource-constrained, low-power embedded system. In this paper, recursive model-based SoC estimation algorithms, such as the extended Kalman filter, have been identified as well-suited solutions for implementation on an embedded platform, providing a good compromise between estimation accuracy and computational complexity that is correlated to energy consumption. Assuming that a decrease in the estimation rate will result in a decrease in both accuracy and energy consumption of the estimator, a theoretical analysis has been conducted to establish how these two metrics depend on the estimation rate. To verify results obtained in theory, two extended Kalman filter-based SoC estimation algorithms of different complexities have been implemented and compared in terms of accuracy, quantified by root mean square error (RMSE), and energy consumption. The obtained results confirm that for a selected type of recursive model-based SoC estimation algorithm, it is possible to achieve an optimal algorithm estimation rate in the sense of satisfactory accuracy and acceptable energy consumption. The analysis and results presented in this paper establish a foundation for a future development of energy-efficient algorithms for SoC estimation in applications where the energy consumption of the estimation process is comparable to the energy consumption of the complete system.

Magnetoresistive angle position sensors are, beside Hall effect sensors, especially suitable for usage within servo systems due to their reliability, longevity, and resilience to unfavorable environmental conditions. The proposed distributed method for self-calibration of magnetoresistive angular position sensor uses the data collected during the highest allowed speed shaft movement for the identification of the measurement process model parameters. Data acquisition and initial data processing have been realized as a part of the control process of the servo system, whereas the identification of the model parameters is a service of an application server. The method of minimizing of the sum of algebraic distances of the sensor readings and the parametrized model is employed for the identification of parameters of linear compensation, whereas the average shaft rotation speed has been used as a high precision reference for the identification of parameters of harmonic compensation. The proposed method, in addition to a fast convergence, provides for the increase in measurement accuracy for an order of magnitude. Experimentally obtained measurement uncertainty was better than 0.5°, with the residual variance less than 0.02°, comparable to the sensor resolution.

This paper presents a time-domain analysis of an N-bit pulse-width-modulated digital-to-analog converter (PWM DAC) employing a first-order passive resistor-capacitor (RC) filter. Exact analytical expressions are derived for the output settling time in both rising and decreasing digital-code transition modes. The worst-case condition is identified, and the settling-time criterion is expressed as a function of the DAC resolution N, tolerance ε, and normalized filter parameter k = T/RC. The derived criterion is compared with a commonly used first-order RC settling approximation. For N = 8 and ε = 1/4, the proposed worst-case criterion gives a discrete settling interval of 823 PWM periods, whereas the literature-based estimate gives 355 periods. The analytical results are confirmed by numerical evaluation and LTspice transient simulations and are further supported by experimental measurements obtained using a microcontroller-based PWM generator and a passive RC filter. The results confirm the duty-cycle dependence of the steady-state ripple and demonstrate that the proposed criterion provides a conservative design rule for selecting PWM DAC parameters while balancing accuracy, ripple, and settling speed.