Sensor fusion has gained a great deal of attention in recent years. It is used as an application tool in many different fields, especially the semiconductor, automotive, and medical industries. However, this field of research, regardless of the field of application, still presents different challenges concerning the choice of the sensors to be combined and the fusion architecture to be developed. To decrease application costs and engineering efforts, it is very important to analyze the sensors’ data beforehand once the application target is defined. This pre-analysis is a basic step to establish a working environment with fewer misclassification cases and high safety. One promising approach to do so is to analyze the system using deep neural networks. The disadvantages of this approach are mainly the required huge storage capacity, the big training effort, and that these networks are difficult to interpret. In this paper, we focus on developing a smart and interpretable bi-functional artificial intelligence (AI) system, which has to discriminate the combined data regarding predefined classes. Furthermore, the system can evaluate the single source signals used in the classification task. The evaluation here covers each sensor contribution and robustness. More precisely, we train a smart and interpretable prototype-based neural network, which learns automatically to weight the influence of the sensors for the classification decision. Moreover, the prototype-based classifier is equipped with a reject option to measure classification certainty. To validate our approach’s efficiency, we refer to different industrial sensor fusion applications.

The semiconductor industry is strongly increasing the production capacities and the product portfolio for a wide range of applications that are needed in the worldwide supply chains e.g. the automotive, computer and security industry. The complex manufacturing processes require more automation, dig- italisation and IoT frameworks, especially for highly automated semiconductor manufacturing plants. Over the last years, this industry spent much effort to control highly sensitive materials in production by product monitoring using advanced process control by various sensors in production. Nevertheless, until today, sensor integration, especially for such sensors that are not supported by the equipment vendors, is time-consuming and complicated. This article aims to use a micro-service-based approach by Eclipse Arrowhead as an open-source microservice architecture and implementation platform [1]. This architecture is an easy and powerful framework that can be used for multiple sensor applications to control the manufacturing material flow in a modern semiconductor plant with a high product mix. The article describes how the engineering process was designed, the architecture of the use case and the main benefits in the operational business are shown.

Industry 4.0 has become a general keyword over the last years. It is based on the inclusion of automation by increasing connectivity in various tasks during the production process. This fact did not exclude the human's effort whose presence remains important, especially the interaction between humans and robots will be a key element in the future manufacturing. In automated production lines, we find both humans and robots operating side-by-side in hybrid workplaces. The major focus for this workplaces today and in the future is to establish a safe work environment. However, what if safety meets "collaborative efficiency"? The system presented in this paper relies on the fusion of data coming from a Time of Flight (ToF) sensor and a 60 GHz radar sensor. The data are analyzed and evaluated using deep learning (DL) algorithms. The purpose is to detect humans and track their movements in the observed area. The resulted perception system can be installed somewhere in a room or on a moving system. A first demonstrator has been developed, tested and evaluated. An additional graphical interface was developed to show in real time the capability of the data fusion system. The system can detect up to 5 persons in a selected area with 98% confidentiality. The so-described system is able as well to estimate each person DoM and the person's instantaneous speed and position. Based on the output of our developed system, it is possible to define industrial use cases as well as many other different applications in different fields.

No abstract

No abstract