Automated Guided Vehicles (AGVs) have become indispensable for industrial and warehouse applications, but they still face numerous challenges in real-world environments. In particular, interaction with humans places stringent demands on 3D sensing and processing systems. Safety is paramount, as autonomous AGVs must not pose any risk to human workers.

Integrating artificial intelligence offers significant potential for improvement. Combined with reliable emergency stop systems, AI can assess the surrounding environment and react to real-world situations. There are two main approaches to AI-based data processing. The first relies on centralized cloud computing, but latency prevents its use for safety-critical functions. The second is local processing on the vehicle itself, which limits scalability and efficiency in multi-robot systems. Especially in dynamic industrial environments, there is a lack of flexible, distributed systems capable of sharing computational workloads across multiple devices.

Safe and ready for collaboration

To address this gap, the researchers involved in the PLATON project developed an integrated platform in which each autonomous vehicle is equipped with multiple spatially resolving radar sensors. These sensors measure distances to surrounding objects in three dimensions, enabling both obstacle detection and precise vehicle localization. In addition, the same radar frontend is used to capture high-resolution synthetic aperture radar (SAR) images.

This approach allows safety-critical and time-sensitive signal processing to take place directly within the sensor. Information about detected obstacles can then be shared with other vehicles in the fleet. More demanding tasks, such as collaborative environment mapping, can be distributed across multiple systems. As a result, the platform scales naturally with the fleet: every additional AGV contributes computing resources that can be allocated dynamically where they are needed most.

A digital twin of the complete system and its sensor technology further accelerates the deployment of new AGV fleets by enabling a virtual map of the operating environment to be generated before the physical vehicles are commissioned.

A heart beating at radar frequencies

To make all of this possible, the Fraunhofer Institute for Reliability and Microintegration IZM joined forces with the Fraunhofer Institute for High-Frequency Physics and Radar Technology FHR to develop hybrid radar hardware capable of performing two tasks simultaneously: measuring distances in all spatial directions and generating radar images. Two synchronized radar chips form the heart of the system, mounted on a printed circuit board (PCB) that also integrates the hardware required for both local and networked processing functions. The capabilities and features of the platform have already been demonstrated in a prototype vehicle. Together with the prototypes developed by the project partners, some of which offer enhanced interaction with their surroundings, it provides a glimpse of how collaboration between humans and autonomous robots could look in the future.

The project “Distributed Computing Platform for Radar-Based 3D Environment Sensing in Safe Autonomous Driving” (PLATON) ran from 1 November 2022 to 30 April 2026 and was carried out by Pilz GmbH & Co. KG as project coordinator, together with Creonic GmbH, let's dev GmbH & Co. KG, Reeb Engineering GmbH, OFFIS e.V., and Fraunhofer IZM. The project was funded under the Federal Ministry for Research, Technology, and Space's funding programme “Electronic Systems for Trusted and Energy-Efficient Distributed Computing in Edge Computing” (OCTOPUS) under funding code 16ME0750.

(Text: Steffen Schindler)