Issue 41 Unmanned Systems Technology December/January 2022 PteroDynamics X-P4 l Sense & avoid l 4Front Robotics Cricket l Autonomous transport l NWFC-1500 fuel cell l DroneX report l OceanScout I Composites I DSEI 2021 report

37 limited adoption in smaller UAVs owing to their higher cost and weight penalty. A visual SAA system allows a UAV to look in all directions all the time; its coverage is limited only by the number of cameras and computing power the UAV can accommodate. ADS-B is a radio signal sent by aircraft to provide identification and location. Anyone with an ADS-B receiver can see those cooperative planes on a map, and an autonomous UAV can make an immediate decision about whether an avoidance manoeuvre is necessary. However, ADS-B may sometimes be insufficient if used only by itself, since most manned aircraft and some unmanned aircraft haven’t yet adopted ADS-B transmitters. However, a European Union project called EUDAAS is planning to develop and validate a European detect & avoid (DAA) system over the next three years that would allow large military remotely piloted aircraft systems (RPASs) to operate in European air traffic. This DAA system will give the remote pilot the ability to ‘see and avoid’, and will include a fully automatic collision avoidance function that will initiate manoeuvres to avoid collisions with other aircraft if necessary. The system is to be integrated into the European ATM system for manned aircraft, highlighting the need for reliable radio comms. The system will be tested on several unmanned platforms including the Medium-Altitude Long-Endurance RPAS and tactical UAVs, as well as the UMS Skeldar V-200 unmanned helicopter. The technology is intended to be applicable to civil systems and will comply fully with civil requirements. Visual detection Advances in camera sensors, vision algorithms and graphics processing units (GPUs) are already boosting the performance of visual sense & avoid systems, but still with the challenge of minimising the SWaP. For example, the latest visual monitoring system uses an 8.9 MP sensor to provide 80 º horizontal and 50 º vertical fields of view that can detect another small aircraft at an average range of 1200 m. This is achieved using machine learning and computer vision algorithms trained with 600 Tbytes of data. The sensor is housed in a module that measures 60 x 60 x 105 mm, while the processing is handled by a GPU in a module measuring 77 x 110 x 36 mm. Together, these weigh 482 g and have a typical power consumption of 10 W, 15 W peak. But this is a single sensor and so can monitor in only one direction, usually forwards. The same machine learning algorithms are used with five cameras to provide a 360 º horizontal field of view with the same 50 º vertical view. This delivers the same 1200 m range in a larger module that measures 103 x 168 x 48 mm. The cameras are connected to the module via Gigabit Ethernet and so can be mounted on various parts of the airframe to provide an unobstructed 360 º view. However, this comes at a cost in the SWaP, as the extra cameras and larger module weigh 2.4 kg and consume an average of 45 W with a peak of 60 W. This limits the use of such systems to larger UAVs that can carry the additional weight and do not rely on battery power. It becomes even more of a challenge when integrating other sensors for SAA. For a collision avoidance system, various sensors have to be taken into account, such as a traffic alert and collision Sense & avoid systems | Focus Unmanned Systems Technology | December/January 2022 A long-range unmanned helicopter is being used in the EUDAAS project to test sense & avoid technologies (Courtesy of UMS Skeldar)