Rutronik System Solutions develops demo for AI control

The latest example of this is an AI control system with gesture recognition using RAB3 radar and keyword recognition based on Infineon's PSOCTM Edge E84. This microcontroller enables decentralised calculation of neural networks ‘on the Edge’, thus increasing the speed, stability, and security of the developed system.

Accordingly, the PSOCTM Edge E84 and the associated evaluation kit form the core – the ‘brain’ – of the demo designed by Rutronik System Solutions. Due to the enormous volume of data that inevitably needs to be processed in AI applications, these usually use Cloud servers for the calculations. The integration of the PSOCTM Edge E84 now brings this processing power from the Cloud to the Edge, creating a closed system. This capability, combined with state-of-the-art security features such as the lockstep-secured enclave in a low-power, always-on domain, the Infineon Edge Protect Category 4 and the standard Trusted Firmware-M activation and Mbed-TLS for crypto operations, ensures maximum security.

The Arm Cortex-M microcontrollers of the PSOC Edge E84 series offer high-performance, secure MCUs with low power consumption and hardware-supported ML acceleration. The PSOCTM Edge devices are based on the powerful Cortex-M55, including Helium DSP support in combination with Arm Ethos-U55, as well as a Cortex-M33 with Infineon's ultra-low-power NNLite hardware accelerator, which powers neural networks in machine learning and AI applications.

With the ‘eyes’ of Rutronik System Solutions

The demo learns to see through the Rutronik Adapter Board RAB3 – Radar, which is plugged into the Evaluation Kit via the Arduino interface. It features the Infineon XENSIV 60GHz radar sensor, a discrete radar IC, and thus the latest radar technology. The advantages of radar sensors include the fact that they can be installed behind certain materials. This makes them invisible and makes it easier to clean the surfaces. This is particularly advantageous for environments with high hygiene requirements. Furthermore, radar is ideal for outdoor applications because the technology works stably and reliably even in sunlight. The XENSIV also impresses with its small form factor and low power consumption. It has one transmitting and three receiving antennas. The L-shaped antenna enables array, horizontal and vertical angle measurements, thus allowing comprehensive radar scanning.

Time savings for developing your own designs

The training process for gesture and voice-controlled applications is extremely time-consuming, but it is essential for the optimal functionality of the system. Rutronik System Solutions has already taught the MCU which meanings are to be assigned to a wide range of hand movements and which reaction should follow predefined keywords (Down, Go, Left, No, Right, Stop, Up and Yes). In order to also receive a reaction to movements that are less precisely executed, the developers have dispensed with the use of a neural network at this point. This means that the demo is ready for the specific customer application. Of course, individual gesture control commands can still be taught. If an application requires a higher level of robustness in terms of gesture control, it is possible to achieve this by integrating a neural network. They are also indispensable if a specific true/false classification of actions is desired.

In addition, the PSOC Edge E84 features always-on acoustic activity detection, which enables HMI operation with low active and standby power consumption, thus extending the battery life of battery-powered products. For the keyword demo, these are connected to digital microphones from Infineon, effectively acting as the ‘ears’ for near real-time voice control of the motor.

The demo's motor control is implemented via an eight-pole BLDC motor with integrated HALL sensors, also from Infineon. This runs in parallel to the processes for speech recognition and gesture control on the same microcontroller.

Rutronik System Solutions complements the demo with a 1024 x 600 IPS TFT LCD touch display from Raystar, which is optimally supported by Infineon's PSOC Edge E84 thanks to its 2.5D graphics processing unit with low power consumption and the display controller with MIPI-DSI/DBI interface (up to 1024 x 768).

Systematically optimised for the Edge

With the shift of AI away from the Cloud, Rutronik System Solutions offers the opportunity to significantly optimise smart (outdoor) applications.

“The systems that can be implemented with our demo are significantly less susceptible to interference from sunlight thanks to radar, more secure against data tapping thanks to the 'on the Edge' approach, and, above all, we are able to reduce the latency times that pose a significant problem for efficient processes in conventional Cloud solutions,” explains Stephan Menze, Head of Global Innovation Management at Rutronik. “With this demo, which – like all developments from Rutronik System Solutions – relies on the integration of the most innovative components from our partners, we enable interested customers to implement their own smart applications in a fraction of the actual development time.”