Volume 90, Issue 3 - Special Issue: Measurement Systems and Sensors with Cognitive Features/Messsysteme und Sensoren mit kognitiven Funktionen; Guest Editors: Klaus-Dieter Sommer, Michael Heizmann and Andreas Schütze

Measurement data can be considered complete only with an associated measurement uncertainty to express knowledge about the spread of values reasonably attributed to the measurand. Measurement uncertainty also allows to assess the comparability and the reliability of measurement results as well as to evaluate decisions based on the measurement result. Artificial Intelligence (AI) methods and especially Machine Learning (ML) are often based on measurements, but so far, uncertainty is widely neglected in this field. We propose to apply uncertainty propagation in ML to allow estimating the uncertainty of ML results and, furthermore, an optimization of ML methods to minimize this uncertainty. Here, we present an extension of a previously published automated ML toolbox (AMLT), which performs feature extraction, feature selection and classification in an automated way without any expert knowledge. To this end, we propose to apply the principles described in the “Guide to the Expression of Uncertainty in Measurement” (GUM) and its supplements to carry out uncertainty propagation for every step in the AMLT. In previous publications we have presented the uncertainty propagation for some of the feature extraction methods in the AMLT. In this contribution, we add some more elements to this concept by also including statistical moments as a feature extraction method, add uncertainty propagation to the feature selection methods and extend it to also include the classification method, linear discriminant analysis combined with Mahalanobis distance. For these methods, analytical approaches for uncertainty propagation are derived in detail, and the uncertainty propagation for the other feature extraction and selection methods are briefly revisited. Finally, the use the uncertainty-aware AMLT is demonstrated for a data set consisting of uncorrelated measurement data and associated uncertainties.

Explainable artificial intelligence (XAI) can make machine learning based systems more transparent. This additional transparency can enable the use of machine learning in many different domains. In our work, we show how XAI methods can be applied to an autoencoder for anomaly detection in a sensor-based sorting system. The setup of the sorting system consists of a vibrating feeder, a conveyor belt, a line-scan camera and an array of fast-switching pneumatic valves. It allows the separation of a material stream into two fractions, realizing a binary sorting task. The autoencoder tries to mimic the normal behavior of the nozzle array and thus can detect abnormal behavior. The XAI methods are used to explain the output of the autoencoder. As XAI methods global and local approaches are used, which means we receive explanations for both a single result and the whole autoencoder. Initial results for both approaches are shown, together with possible interpretations of these results.

Artificially intelligent automation has not only impact on sensor technologies, but also on comprehensive multiple sensor systems for assisting situational awareness and decision-making. This is particularly true for integrated Manned-unManned-Teaming (MuM-T), for example. From a systems engineering perspective which does not exclude applications in the defence domain, three tasks need to be fulfilled: (1) Design artificially intelligent automation in a way that human beings are mentally and emotionally able to master each situation. (2) Identify technical design principles to facilitate the responsible use of AI in ethically critical applications. (3) Guarantee that human decision makers always have full superiority of information, decision-making, and options of action. Our discussion of AI-driven systems for multiple sensor data fusion results in recommendations and key results. We are addressing the algorithms needed, the data to be processed, the programming skills required, the computing devices to be used, the inevitable anthropocentric design, the reviewing of R & D efforts necessary, and the integration of different dimensions in a systems-of-systems point of view.

In this paper, we present a collection of machine learning assisted distributed fiber optic sensors (DFOS) for applications in the field of infrastructure monitoring. We employ advanced signal processing based on artificial neural networks (ANNs) to enhance the performance of the dynamic DFOS for strain and vibration sensing. Specifically, ANNs in comparison to conventional and computationally expensive correlation and linearization algorithms, deliver lower strain errors and speed up the signal processing allowing real time strain monitoring. Furthermore, convolutional neural networks (CNNs) are used to denoise the dynamic DFOS signal and enable useable sensing lengths of up to 100 km. Applications of the machine learning assisted dynamic DFOS in road traffic and railway infrastructure monitoring are demonstrated. In the field of static DFOS, machine learning is applied to the well-known Brillouin optical frequency domain analysis (BOFDA) system. Specifically, CNN are shown to be very tolerant against noisy spectra and contribute towards significantly shorter measurement times. Furthermore, different machine learning algorithms (linear and polynomial regression, decision trees, ANNs) are applied to solve the well-known problem of cross-sensitivity in cases when temperature and humidity are measured simultaneously. The presented machine learning assisted DFOS can potentially contribute towards enhanced, cost effective and reliable monitoring of infrastructures.

End-of-line (EOL) quality assurance of finished components has so far required additional manual inspections and burdened manufacturers with high labor costs. To automate the EOL process, in this paper a fully AI-based quality classification system is introduced. The components are automatically placed under the optical inspection system employing a robot. A Convolutional Neural Network (CNN) is used for the quality classification of the recorded images. After quality control, the component is sorted automatically in different bins depending on the quality control result. The trained CNN models achieve up to 98.7% accuracy on the test data. The classification performance of the CNN is compared with that of a rule-based approach. Additionally, the trained classification model is interpreted by an explainable AI method to make it comprehensible for humans and reassure them about its trustworthiness. This work originated from an actual industrial use case from Witzenmann GmbH. Together with the company, a demonstrator was realized.

Artificial intelligence methods based on machine learning or artificial neural networks have become indispensable in camera-based driver assistance systems, and also represent an essential building block for future autonomous driving. However, the great successes of these evaluation methods in environment perception and also driving planning are accompanied by equally great challenges in the validation and verification of these systems. One of the essential aspects for this is the required guaranteed safety of the functions under mass production conditions of the vehicles. This article explains this point of view using a detailed example from the field of camera-based driver assistance systems: the determination of inspection limits at the end of the production line. The camera is one of the most important sensor modalities for vehicle environment sensing and as such, the quality of the camera systems plays a key role in the safety argumentation of the overall system. Several illustrative application examples (role of simulations, calibration, influence of the windshield) will be presented. The basic ideas presented can be well transferred to the other sensor modalities (lidar, radar, ToF, etc.). The investigations/evidence show that doubts are allowed whether or how fast autonomous driving on level L4/5 will take hold as robotaxis or – even more challenging – in private ownership on a larger scale.