34 That typical six hours of extra endurance for the non-VTOL configuration can be very significant, Mendes stresses. “We see a lot of systems that are just VTOL, and that’s a tremendous limitation on the system. So, we thought the best of both worlds would be to have the option of being either pure fixed-wing or pure VTOL.” Beacon landings The VTOL capability also brings the potential for precision landings in confined spaces, difficult terrain and onto moving vessels. In VTOL mode, the AR3 Mk9 needs only five square meters of relatively flat surface to land on, whether that’s on a ship’s deck, a mountain ridge or even the roof of a building. The standard approach – relying solely on GPS – can provide reasonable accuracy under ideal conditions, but in GNSS-denied or contested environments, that’s simply not an option, says Nunes. This is where the company’s Lidar-beacon-based landing system, supported by onboard cameras, comes into play. The aircraft first navigates to a rough landing position, then detects and locks onto the beacon, guiding itself down with precision. To land on a moving vessel, the aircraft matches its forward speed with the ship during descent, achieving zero relative velocity, enabling landings with sub-meter accuracy, even in rough seas, he explains. Mendes notes that this precision landing system represents an ongoing, multidisciplinary initiative that brings together teams from airframe design, payload integration, artificial intelligence (AI), and data processing and autopilot development. The system can use visual aids from onboard cameras, Lidar, RF positioning and GPS/visual GPS, he says, processed in real time to build a precise relative position fix. Aided by AI, data fusion happens at the hardware level, with a dedicated computing module weighing inputs from each sensor based on environmental conditions and mission requirements. The output of the fusion process is the ‘best’ estimate of the vehicle’s state, which the autopilot uses to control the landing. “These best estimates are based on all the inputs you have, and machine learning is really important because you need to consider past experience based on those inputs,” he says. “It’s not a direct application of rules.” Threat avoidance As well as the precision landing system, AR3 Evolution’s sensors also contribute more general situational awareness functions needed for beyond-visualline-of-sight operations, including the avoidance of obstacles, other air traffic and, in war zones, threats. Of the several technologies available, Tekever is most advanced with visual-based detection algorithms for AR3. Nunes says that AR3’s cameras look all around, not just forward, because of the threat from hostile drones sneaking up on it, either to engineer a collision or to launch a weapon at it. “We have already deployed mechanisms to avoid that, using visual detections and evasive manoeuvres. We are working with acoustic and radar technologies as well for detect-and-avoid solutions for ‘noncooperative’ airspace users.” This is in addition to the integration of collaborative systems including ADS-B and transponder signals, he says. In-house autopilot The autopilot itself is an in-house development. Throughout the entire programme, as a matter of course, they were constantly asking themselves whether it would be better to buy an existing component or subsystem or to develop their own. They went down the in-house development route with the autopilot because they needed some very specific behaviours, and being vertically integrated made it more efficient for them to bring in their own multidisciplinary team to do the job. “Having this ecosystem working together to deliver a system brings much more value than just going out to the market and buying something,” he emphasises. The decision also reflected where Tekever had come from. “We were a software company way before we started building aircraft and so, for us, it was only natural to build the intelligence engine of the system – the autopilot.” Returning to specific behaviours, he explains that to succeed with deep integration of visual-based navigation or other alternative means of location awareness – and to do swarming, for Dossier | Tekever AR3 August/September 2025 | Uncrewed Systems Technology In managing the Evolution’s radar cross section (RCS), designers have given it a small frontal area, but have had to eschew edge alignment with the V-tail and accept sensors that increase the RCS
RkJQdWJsaXNoZXIy MjI2Mzk4