embedded world 2024

Edge AI Inference

Our embeddif.[ai] technology enables the implementation of AI models on embedded sensor modules for vibration, speech and audio analysis without a cloud connection. Through embeddif.[ai]'s resource-efficient inference, AI applications can be executed directly on the hardware of sensor modules. This enables the identification and analysis of sensor signals with low energy consumption in the mW range on commercial microcontrollers.

Start Video on Youtube

On-Device Training

With our embeddif.[ai] products, you can realize true self-learning IoT systems. The training of AI models takes place directly at the sensor, eliminating the need for cloud communication. This opens new possibilities for self-learning AI applications in the IoT domain. Our hardware-optimized training methods even allow for resource-efficient training of CNNs on embedded systems. This makes autonomous computer vision applications easily achievable.

more info

Extreme Edge Computing

Low latency, high data throughput, absolute data sovereignty are advantages when AI inferences are processed directly on end devices. But computing power and storage capacity are often limited there. With our innovative approach, neural networks are therefore executed in a self-regulated manner on multiple IoT devices. Our demonstrator visualizes the data transfer and shows how your own handwriting is processed in a distributed computing process.

more info

Neuromorphic Computing for Edge AI

Machine Learning applications in embedded devices are a strongly emerging trend. A large number of AI chips has been announced, the first products for embedded AI are on the market. Current neural network architectures like deep neural networks require high computational complexity and power consumption. Neuromorphic hardware in contrast relies on massive parallel processing and performs calculations, e.g. for Machine Learning, faster and with less power. Efficient architectures for neuromorphic hardware with respect to computational performance, power consumption and chip area are therefore a key element for a widespread deployment of neural networks in embedded devices.

more info

JPEG XS SDK (Software Development Kit)

JPEG XS is a video codec for all CPU-, GPU-, FPGA-based or Nvidia Jetson designs which call for a transfer of high resolution video data and an ultra-low latency accordingly to IPMX and ST2110-22.. Fraunhofer IIS provides a JPEG XS SDK Application Support Package to easily implement JPEG XS in typical workflows like camera to stream and stream to display.

The JPEG XS SDK is available under licensing. Ask for an evaluation kit or more details.

more info

Energy autarkic IoT sensors and Energy Harvesting

With energy harvesting technologies, we generate electricity from the smallest amounts of light, slight temperature differences or barely perceptible vibrations and operate IoT sensors in an energy autarkic manner. A new feature is an energy prediction algorithm that uses recorded energy yield data from an energy harvester to estimate what the future yield will be.

MORE INFO

Class B and C: New Features for Wireless mioty®-LPWAN

With the new Class B, control commands can also be sent from a base station to the devices. Broadcast features allow all actuators to be controlled via downlink or multicast in a group of actuators. Class C enables immediate downlink communication for fast activation, as demonstrated by a wireless lighting control system.

MORE INFO

RFicient® Ultra-low-power Wake-Up-receiver

The energy-efficient RFicient® radio receiver technology with a current consumption less than 3 µA can receive radio telegrams in three license-free frequency bands simultaneously and independently of each other. This makes positioning systems for the localization of objects in warehouses possible, but also worldwide mobile applications, such as container tracking without manual frequency switching. RFicient® is also suitable for building automation, smart lighting, electronic shelf labels, remote maintenance and control, and wireless sensor networks in general. Years of maintenance-free operation is possible with very small batteries and solar cells or with energy harvesting. The particularly short response time in milliseconds also enables real-time applications.

more info

IC-Design

As one of the leading IC design facilities in Europe, we develop customized solutions to meet the constantly evolving requirements of industrial applications. Our work being independent of technology and manufacturer, we are able to provide our customers with optimum solutions. Our focus is on development of mixed-signal ASICs, intelligent integrated sensor systems as well as on design solutions for increasingly complex electronic systems. Whether for industrial, communications, automotive or any other applications, we are your partner from the idea to the series production.

more info

embedded world 2024

Nürnberg Messe / April 09, 2024 - April 11, 2024