Secure Human-Robot Collaboration with Neuromorphic Sensor Platform

[Erkan Teskancan]

## Safe Human-Robot Collaboration with Neuromorphic Sensor Platform

Fraunhofer Institutes have developed a new neuromorphic sensor platform called NeurOSmart, which offers energy efficiency and real-time performance in industrial robotic applications. This platform integrates LIDAR sensors, embedded artificial intelligence, and neuromorphic chips to enhance safety in human-robot collaboration.

NeurOSmart enables robots to continuously monitor human presence and adapt in real-time while performing complex tasks. This provides opportunities for safer and more effective collaboration in manufacturing environments.

### Integrated 3D Sensing and MEMS LIDAR

• At the heart of NeurOSmart is a LIDAR system that observes the workspace from a bird's-eye view.
• It emits short laser pulses in the near-infrared wavelength and measures 3D distances through reflections.
• Movable MEMS mirrors scan the laser across the entire workspace, generating high-resolution spatial data.
• The mirrors are made from piezoelectric aluminum scandium nitride (AlScN) material, approximately 1 micrometer thick, which ensures high scanning performance with low energy consumption.

### Edge Data Processing with Embedded Artificial Intelligence

• Data processing on the NeurOSmart platform occurs directly within the sensor system.
• Instead of sending large image data to central control units, pre-processing is done at the edge.
• AI algorithms developed by Fraunhofer IMS collect incoming signals and identify areas of interest.
• This method reduces data rates and energy consumption while focusing resource utilization on human-robot interaction.
• AI-supported evaluation is integrated into the sensor hardware, minimizing the delay between sensing and reaction.

### Millisecond Response Times with Neuromorphic Computing

• For deeper analysis and motion control decisions, a neuromorphic accelerator chip developed by Fraunhofer IPMS is used.
• This chip, inspired by biological neural networks, consists of numerous interconnected processing units.
• It enables parallel and energy-efficient signal evaluation.
• In conjunction with AI models from Fraunhofer IAIS, it processes sensor input and initiates mechanical reactions within a few milliseconds.
• This allows heavy industrial robots to immediately slow down or stop when a human enters a defined safety zone.
• Unsafe scenarios in production environments are subjected to AI training in simulation environments developed by Fraunhofer IAIS.

### Modular Platform for Industrial Applications

• The NeurOSmart platform is a standardized modular technology platform led by Fraunhofer ISIT, with participation from IPMS, IMS, IWU, and IAIS.
• Industrial partners can adapt the platform to their needs.
• LIDAR sensing, AI-supported edge processing, and neuromorphic chips can also be used in energy-efficient autonomous systems (drones, agricultural monitoring).
• It is an ideal solution for low-latency decision-making and long battery life.

NeurOSmart combines sensor hardware, artificial intelligence, and brain-inspired computing architecture to enable safe and fast human-robot collaboration in modern production environments.