Sensor Calibration and Hysteresis Compensation With Heteroscedastic Gaussian Processes

Urban, Sebastian; Ludersdorfer, Marvin; Smagt, Patrick van der

doi:10.1109/jsen.2015.2455814

Cited by 23 publications

(27 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This section verifies the proposed DVSHGP and SVSHGP against existing scalable HGPs on a synthetic dataset and four real-world datasets. The comparison includes (i) GPVC [8], (ii) the distributed variant of PIC (dPIC) [29], (iii) FITC [48], and (iv) the SoD based empirical HGP (EHSoD) [11]. Besides, the comparison also employs VSGP [27], a homoscedastic GP which can be used to showcase the benefits brought by the consideration of heteroscedasticity.…”

Section: Numerical Experimentsmentioning

confidence: 99%

“…The developments in scalable homoscedastic GPs have thus motivated us to scale up HGPs. Alternatively, we could combine the simple Subset-of-Data (SoD) approximation [45] with the empirical HGP [11] which trains two separate GPs for predicting mean and variance, respectively. The empirical variance however is hard to fit since it follows an asymmetric Gaussian distribution.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Large-Scale Heteroscedastic Regression via Gaussian Process

Liu

Ong

Cai

2021

IEEE Trans. Neural Netw. Learning Syst.

View full text Add to dashboard Cite

Heteroscedastic regression which considers varying noises across input domain has many applications in fields like machine learning and statistics. Here we focus on the heteroscedastic Gaussian process (HGP) regression which integrates the latent function and the noise together in a unified nonparametric Bayesian framework. Though showing remarkable performance, HGP suffers from the cubic time complexity, which strictly limits its application to big data. To improve the scalability of HGP, we first develop a variational sparse inference algorithm, named VSHGP, to handle large-scale datasets. Furthermore, two variants are developed to further improve the scalability and capability of VSHGP. The first is stochastic VSHGP (SVSHGP) which derives a relaxed evidence lower bound factorized over data points, thus enhancing efficient stochastic variational inference. The second is distributed VSHGP (DVSHGP) which (i) follows the Bayesian committee machine formalism to distribute computations over multiple local VSHGP experts with many inducing points; and (ii) adopts hybrid parameters for experts to guard against over-fitting and capture local variety. Superiority of DVSHGP and SVSHGP as compared to existing scalable heteroscedastic/homoscedastic GPs is then verified using a synthetic dataset and four real-world datasets.

show abstract

Section: Numerical Experimentsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Large-Scale Heteroscedastic Regression via Gaussian Process

Liu

Ong

Cai

2021

IEEE Trans. Neural Netw. Learning Syst.

View full text Add to dashboard Cite

show abstract

“…As the operating supply voltage is 1.2 V, the drain-source voltage V dsc of the NMOSFET M nc is greater than 100 mV, thus it is working in the subthreshold-saturation region with a total current value of (N + 1)I 1 . Hence, using Equations (13) and (14) in Equation (15) results in:…”

Section: Temperature Sensing Corementioning

confidence: 99%

“…One of the common post-processing methods for either BJT-based or MOSFET-based temperature sensor designs is the calibration, which is required to assure that the measurements are accurate and within the specifications [12][13][14]. In the aforementioned literature, either the single-point or the two-point calibration method is used to adjust the attribute of the temperature sensor.…”

Section: Introductionmentioning

confidence: 99%

A 40 nW CMOS-Based Temperature Sensor with Calibration Free Inaccuracy within ±0.6 °C

Chouhan

Halonen

2019

Electronics

View full text Add to dashboard Cite

In this study, a temperature equivalent voltage signal was obtained by subtracting output voltages received from two individual temperature sensors. These sensors work in the subthreshold region and generate the output voltage signals that are proportional and complementary to the temperature. Over the temperature range of −40 ∘C to +85 ∘C without using any calibration method, absolute temperature inaccuracy less than ±0.6 ∘C was attained from the measurement of five prototypes of the proposed temperature sensor. The implementation was done in a standard 0.18 μ m CMOS technology with a total area of 0.0018 mm 2. The total power consumption is 40 nW for a supply voltage of 1.2 V measured at room temperature.

show abstract

“…Relevant works on this scope have been developed by Lebosse [9], Schofield [5], Dabling [12], Vecchi [13] and Urban [14]. Likewise, authors' previous work has demonstrated that the A201 sensors, working on the piezoresistive principle, are also capable of exhibiting a piezocapacitive response.…”

Section: Introductionmentioning

confidence: 99%

Experimental characterization, modeling and compensation of hysteresis in force sensing resistors

Paredes-Madrid

Matute

Cruz-Pacheco

et al. 2018

DYNA

View full text Add to dashboard Cite

Force Sensing Resistors (FSRs) exhibit considerable amounts of hysteresis and repeatability error inhibiting their usage in applications that require high-accurate force readings. This paper presents the hysteresis characterization and modeling of the Tekscan A201-1 FSR employing the Preisach Operator (PO) function. In order to compensate for hysteresis during sensor operation, the inverse PO was numerically found on the basis of the Closest Match Algorithm (CMA). A test bench, capable of handling sixteen sensors simultaneously, was built, which allowed the characterization and later testing of the CMA. Grip force profiles were applied to the sensors during testing and the experimental results showed a considerable reduction in the force estimation error compared with the linear regression method proposed by the manufacturer. These results enable a wider use of FSRs in applications with tight accuracy requirements. Finally, a generalized sensor model for hysteresis compensation that simplifies the obtaining of PO parameters is presented.Keywords: hysteresis; force sensing resistors; Preisach operator; closest match algorithm.Caracterización experimental, modelado y compensación de la histéresis en sensores de fuerza resistivos Resumen Los Sensores de Fuerza Resistivos (FSRs) despliegan cantidades considerables de histéresis y de error de repetitividad que inhiben su uso en aplicaciones que requieren lecturas de fuerza de alta precisión. En este trabajo se presenta la caracterización y modelado de histéresis del sensor de presión Tekscan A201-1 empleando la función Operador de Preisach (OP). Con el fin de compensar la histéresis durante el funcionamiento del sensor, el OP inverso se halló numéricamente sobre la base del algoritmo de coincidencia más cercana (CMA). Se construyó un banco de pruebas, capaz de manejar dieciséis sensores simultáneamente, lo que permitió la caracterización y posterior prueba del CMA. Los perfiles de fuerza de agarre se aplicaron a los sensores durante la prueba y los resultados experimentales mostraron una reducción considerable del error de estimación de la fuerza en comparación con el método de regresión lineal propuesto por el fabricante. Estos resultados abren el camino para un uso más amplio de los FSRs en aplicaciones con exigentes requisitos de precisión. Finalmente, un modelo de sensor generalizado para compensación de histéresis que simplifica la obtención de los parámetros PO, es presentado.Palabras clave: sensores de fuerza resistivos; operador de Preisach; algoritmo de coincidencia más cercana.

show abstract

Sensor Calibration and Hysteresis Compensation With Heteroscedastic Gaussian Processes

Cited by 23 publications

References 17 publications

Large-Scale Heteroscedastic Regression via Gaussian Process

Large-Scale Heteroscedastic Regression via Gaussian Process

A 40 nW CMOS-Based Temperature Sensor with Calibration Free Inaccuracy within ±0.6 °C

Experimental characterization, modeling and compensation of hysteresis in force sensing resistors

Contact Info

Product

Resources

About