Learning-based Energy Consumption Prediction

An accurate soil moisture model is critical in the design and implementation of a smart agriculture system. Accurate soil moisture prediction allows an efficient water resources allocation. This paper presented a soil moisture model using different environmental parameters such as humidity, temperature, light intensity, and rain occurrence as inputs or predictor variables. Gaussian process regression algorithm, a non-parametric machine learning algorithm, was used to develop the model. The most effective kernel function was also determined by developing four different GPR models using a different kernel function. In terms of RMSE, the rational quadratic function obtained the lowest value. To further improve the accuracy of the GPR model, an automated hyperparameter tuning was done using a Bayesian optimization algorithm. Three hyperparameters were tuned using the Bayesian optimization algorithm, which improved the GPR model's performance. The optimized GPR model achieved the lowest RMSE and MAE of 3.596 and 1.176, respectively.

An Optimized Soil Moisture Prediction Model for Smart Agriculture Using Gaussian Process Regression

An accurate soil moisture model is critical in the design and implementation of a smart agriculture system. Accurate soil moisture prediction allows an efficient water resources allocation. This paper presented a soil moisture model using different environmental parameters such as humidity, temperature, light intensity, and rain occurrence as inputs or predictor variables. Gaussian process regression algorithm, a non-parametric machine learning algorithm, was used to develop the model. The most effective kernel function was also determined by developing four different GPR models using a different kernel function. In terms of RMSE, the rational quadratic function obtained the lowest value. To further improve the accuracy of the GPR model, an automated hyperparameter tuning was done using a Bayesian optimization algorithm. Three hyperparameters were tuned using the Bayesian optimization algorithm, which improved the GPR model's performance. The optimized GPR model achieved the lowest RMSE and MAE of 3.596 and 1.176, respectively. 

Hard-processors are those that have their architecture defined from the factory and this makes them less flexible in the face of architecture changes while softprocessors have the advantage of being modifiable in their architecture allowing continuous improvements to be made in the design if they are Open-source Software. In this paper, we propose a design of the first Ecuadorian open-source software softprocessor called RISCV-EC, which is based on a RISC-V single core architecture. In addition, a performance comparison is carried out between the proposed RISCV-EC softprocessor and other processors such as AVR ATMEGA328P, ARM Cortex M1 of the Raspberry Pi Pico and ARM Cortex A9 Zynq-7000 of Xilinx. This comparison consists of estimating the running time for the Fibonacci sequence algorithm increasing the number of iterations from 0 to 500. Owing to the fact that the RISCV-EC softprocessor is implemented in the FPGA Zynq-7000 SoC - Xilinx, same clock was used as the other processors, in such a way that the comparison is a reflection of the architecture improvement and not of an overclock. The Fibonacci series was programmed on all processors using the same Assembler programming language (ASM) to avoid bias in the results by the compilers. Numerical results show that the RISCV-EC softprocessor has a better performance than the ATMEGA328P AVR processor for any given number of iterations of the Fibonacci series. In the case of the ARM Cortex M1 processor, the RISCV-EC processor is better for a number of iterations greater than 300 of the Fibonacci series. Finally, RISCV-EC presents better running time than the ARM Cortex A9 Zynq-700 processor only for a number of iterations lower than 18. 

Softprocessor RISCV-EC for Edge Computing Applications

Electrooculographic techniques are applied in the development of new technologies that compensate for the limitations of people with motor disabilities. The algorithms in charge of classifying these signals play a fundamental role, mainly for Human Computer Interfaces (HCI), specially when the machine learning algorithms are implemented in customized hardware like FPGA. In this work, electrooculography data were collected from 10 healthy subjects during six eye movement tasks. Then, the data were filtered and introduced into supervised and unsupervised learning algorithms with six classification labels. The results obtained showed that the SVM algorithm had 93.5% of accuracy, thus being considered the most efficient of the classification algorithms proposed in this work. Then, we develop a custom hardware architecture for real-time implementation of EOG classification model in al FPGA card. We demonstrate the effectiveness of the proposed framework for EOG data classification.

Electrooculography Signals Classification for FPGA-based Human-Computer Interaction

In this paper, we propose to implement a meteorological monitoring station using embedded systems. This model is possible thanks to different sensors that enable us to measure several environmental parameters, such as i) relative humidity, ii) average ambient temperature, iii) soil humidity, iv) rain occurrence, and v) light intensity. The proposed system is based on a field-programmable gate array device (FPGA). The proposed design aims at ensuring high-resolution data acquisition and at predicting samples with precision and accuracy in real-time. To present the collected data, we develop also a web application with a simple and friendly user interface.

Rebeca Estrada Pico

Papers

FPGA Based Meteorological Monitoring Station