Sick strengthens its activities in the field of AI. To this end, the sensor expert has developed new software applications for logistics automation based on deep learning algorithms.
Using deep learning, sensor systems provide intelligence services for the automatic recognition, testing and classification of objects or features that were previously reserved for humans. This makes Deep Learning - as part of machine learning - one of the most important future technologies in the field of artificial intelligence and at the same time a long-term driver of industry 4.0.
After Sick had already reported at the beginning of the year on the successful implementation of deep learning algorithms in initial pilot projects, a further software application for systems business in logistics automation is now following. The deep learning system recognizes whether a sorter tray in a logistics hub is actually loaded with only one object. This leads to more efficient goods flows.
Neural networks are used to implement deep learning. In contrast to the classical development of algorithms, which is mainly characterized by the manual development of a suitable feature representation, a neural network is trained for optimal features for its task. With application-specific data it is trained again and again in order to adapt to new conditions. Sick uses an independent in-house computer and IT base as the executing unit for the construction of the training data set by capturing and evaluating thousands of images and examples as well as for the training of the neural networks. The complex operations of the deep learning solution for training are calculated on specially equipped computers with high GPU performance. The resulting new deep learning algorithms are made available locally on the sensor and are thus immediately and fail-safe available, for example, on an intelligent camera.
With the implementation of Deep Learning in selected sensors and sensor systems, Sick ignites the next step in the Sick AppSpace ecosystem. Other future products that will work with the new technology and whose customization will generate real added value include image processing sensors and cameras. In principle, however, the concept of the sensor specialized in artificial intelligence can also be applied to simple sensors such as inductive proximity switches, photoelectric reflex switches, ultrasonic sensors and others. In addition, system solutions such as vehicle classification at toll stations offer ample potential for deep learning-based classification of vehicles into toll classes. Sick sensors are already being used here - and with the new deep learning approach, Sick could cut himself another piece of the pie