Diese Website verwendet Cookies, um das Angebot nutzerfreundlicher und effektiver zu machen. Mit der Nutzung dieser Website stimmen Sie der Verwendung von Cookies zu. Weitere Informationen über die Verwendung von Cookies und die Möglichkeit der Verwendung von Cookies zu widersprechen, finden Sie hier.

26. - 28. Februar 2019 // Nürnberg, Germany

Konferenzen und Rahmenprogramm

Zurück zur Tagesansicht
Session 10 I - Autonomous Systems I / Architectures & Applications

olOne: Artificial Intelligence on Chip Vortragssprache Englisch

Embedded systems are at the very heart of most machinery, being used in a wide span of contexts, from industrial processes to consumer appliances. With the prospective growth of Industry 4.0 that promises to enlarge the spectrum of embedded applications towards real-time anomaly detection, predictive maintenance, context awareness and so on, manufacturers will be required to inflate more intelligence into their products. In this paper, we present olOne, the first Artificial Intelligence (AI) technology able to bring real-time raw sensor data interpretation onto tiny micro controllers such as ARM® Cortex® M0, one of the commonly-used processors available on the market. Today, the commercial interest in AI, triggered especially by extensive applications of Deep Learning (DL) and Neural Networks (NN) technology, is hyping. However, there are some technological barriers that generally impede AI rapid adoption in the embedded systems community such as: 1. heavy memory and computational resource requirements considering (NN do not run on chip like ARM® Cortex® M0); 2. need of adequate data scientist skills; 3. additional time spent to produce representative datasets and lead time imposed by the machine learning (ML) training phase 4. reliability and responsiveness, since most current solutions depend on the Cloud Another tricky aspect is the lack of standardization in the modelling process. Machinery builders often have several product lines and the algorithms developed for a product are so customized to the particular application that company know-how is hardly transferred from one engineering team to another (or even between developers of the same team). Our solution is twofold: • implement a lightweight real-time AI engine capable of running on tiny microcontrollers • lower the barrier of AI skills gap by means of a visual integrated development environment (IDE) IDE allows to visually describe sensor data patterns by means of simple drag-and-drop widgets representing trend-, level- or value-related concepts. At the same time, user can enrich the language with new concepts by drawing signal archetypes that can be applied to a variety of waveforms. olOne is integrated into the final application as a C/C++ library (to call in the main loop). It acts as an autonomous software agent that continuously interprets raw data streams (even in very noisy conditions) to effectively detect events designed through the IDE. olOne’s ability to interpret data is inspired to the holonic philosophy as the result of several years of research [1]. Up to authors’ knowledge, there is no other such technology available on the market today. The rest of the paper will be devoted to explain olOne solution in more detail, with particular emphasis on the IDE features and on example use cases. [1] V. Di Lecce, M. Calabrese, C. Martines, From Sensors to Applications: A Proposal to Fill the Gap, Sensors & Transducers Journal, Vol. 18, Special Issue, January 2013, pp. 5-13, ISSN 1726-5479

--- Datum: 27.02.2018 Uhrzeit: 10:00 Uhr - 10:30 Uhr Ort: Conference Counter NCC Ost

Sprecher

top

Der gewählte Eintrag wurde auf Ihre Merkliste gesetzt!

Wenn Sie sich registrieren, sichern Sie Ihre Merkliste dauerhaft und können alle Einträge selbst unterwegs via Laptop oder Tablett abrufen.

Hier registrieren Sie sich, um Daten der Aussteller- und Produkt-Plattform sowie des Rahmenprogramms dauerhaft zu speichern. Die Registrierung gilt nicht für den Ticket- und AusstellerShop.

Jetzt registrieren

Ihre Vorteile auf einen Blick

  • Vorteil Sichern Sie Ihre Merkliste dauerhaft. Nutzen Sie den sofortigen Zugriff auf gespeicherte Inhalte: egal wann und wo - inkl. Notizfunktion.
  • Vorteil Erhalten Sie auf Wunsch via Newsletter regelmäßig aktuelle Informationen zu neuen Ausstellern und Produkten - abgestimmt auf Ihre Interessen.
  • Vorteil Rufen Sie Ihre Merkliste auch mobil ab: Einfach einloggen und jederzeit darauf zugreifen.