Uncrewed Systems Technology 043 l Auve Tech Iseauto taxi l Charging focus l Advanced Navigation Hydrus l UGVs insight l MVVS 116 l Windracers ULTRA l CES 2022 show report l ECUs focus I Distant Imagery

30 Lidar-based localisation as a workaround for the multi-path errors suffered by GNSS when its driverless Renault Zoe taxis traversed the skyscraper-rich streets of Singapore. “That said, we do need an accurate GNSS system for it to serve well as a periodic localisation validator and back- up,” Mossov notes. “We therefore have dual GNSS antennas, outputting accurate heading readings along with position, and the inertial measurement sensor is also integrated into the GNSS system for high-fidelity data sharing and fusion between the two.” Sensor architecture With the Lidar providing localisation and long-range obstacle detection, the cameras serve to sense more immediate threats in 360 º around the vehicle, using AI systems that have been predominantly developed externally and integrated into the Iseauto. They also provide ‘eyes’ for teleoperators if one of them has to step in to analyse a vehicle’s (and its passengers’) safety and wellbeing. “The cameras therefore form a second, tighter safety line around the vehicle, including some that look straight down over the floorspace in front of the doors and the vehicle, because anything there could pose an immediate threat to passengers,” Mossov says. “It may sound odd but it’s also really important if there’s an accident or fault to be able to go back through camera footage and see if, say, there was a rock, some debris or something else in front of the vehicle that you drove over that caused it. Short- and long-run sensing are key for verifying obstacles and how incidents with them can be avoided or mitigated in future behaviours.” The architecture of the sensors around the vehicle’s body is a result of their purpose and consequent fields of view. The Lidars are the most important for long- range detection and collision avoidance in a 180 º field in front of the Iseauto, so three Lidars are installed at the front. Two sit on the roof for a high, long- range perspective (having two also gives good redundancy and ‘second opinion’ validation of object detection), and the other one sits in the bumper to give better detection of objects near or on the ground and as a ‘Lidar of last resort’, as it is the closest to the most likely point of impact in potential collisions from the front. “With the bumper Lidar, object recognition and tracking aren’t so much the key as point density,” Mossov says. “If you suddenly have too many points clustered closely together in one area, it’s because something is coming in fast from the front. “By this logic, we’ve seen that a simple, low-duty algorithm can be relied on to signal an emergency to the main computer, and that an immediate brake is needed. In all other instances, our AI has been programmed in more sophisticated ways to recognise what object it is seeing, where it’s going, and what should be done about it.” The cameras are largely fisheye devices to give the 360 º field of view around the body. As mentioned, these April/May 2022 | Unmanned Systems Technology Dossier | Auve Tech Iseauto Although earlier models used significant quantities of fibreglass, the Iseauto uses largely vacuum-formed ABS for much shorter lead times