Uncrewed Systems Technology 047 l Aergility ATLIS l AI focus l Clevon 1 UGV l Geospatial insight l Intergeo 2022 report l AUSA 2022 report I Infinity fuel cell l BeeX A.IKANBILIS l Propellers focus I Phoenix Wings Orca

8 Platform one Researchers in Germany have developed an autonomous AI system that is comparable to a human driver for detecting pedestrians and other road users (writes Nick Flaherty). The Cognitive Neuroinformatics research group at the University of Bremen has worked with automotive supplier Continental to develop the system as part of the Proreta 5 project. The work includes researchers from TU Darmstadt and the TU Lasi in Romania, and is the latest stage in a project that has been running for 20 years to develop autonomous technologies. The research vehicle was equipped by Continental with sensors and computers to test the resulting functional and verification methods for the automated driving system directly under real conditions. Methods included multi-modal prediction of dynamic behaviour of an object, specifying and testing traffic rules compliance, and logic-based testing to detect unsafe behaviour of AI modules. “The great advantage of AI is that, after a training phase, it is able to draw the right conclusions in unknown situations based on what it has learned,” said Prof Schill, head of the Cognitive Neuroinformatics working group at the University of Bremen. “One part of the project was to observe the human drivers as they reduce and evaluate the complexity of the environment themselves. The adaptive algorithms are now being trained according to similar principles.” In the project, the Cognitive Neuroinformatics group investigated AI methods for recognising objects and obstacles in the environment. An attention-driven pipeline identified relevant areas in camera images using saliency maps that show where a driver’s attention is focused first, for example on other road users or when signs appear. Then the driver’s gaze was projected into the image to expand the relevant area. This distinguished between relevant and non-relevant regions in the image that were used for building mathematically correct models to represent the position, orientation, speed or size of other road users and describe how the other vehicles are moving. The other element of the project was to implement object tracking to perceive road users in the monitoring area and estimate their condition over time using radar and Lidar data. A list of tracked objects is then sent to the prediction, planning and control software modules for further processing, and the state of each object is estimated using a probabilistic filter. Another aspect of the project developed new models to describe articulated vehicles such as buses, trams or vehicles with trailers in a mathematically correct way, and adding this to the tracking algorithms. The first Proreta project in 2002 developed emergency braking systems that are now a mainstream technology. Driverless cars AI rivals human drivers December/January 2023 | Uncrewed Systems Technology The system is now being trained according to how people evaluate a driving environment’s complexity