Unmanned Systems Technology 036

29 Saab Sabertooth AUV | Dossier our autonomy engine that would feed the vehicle with new positions and attitudes, for example,” he says. This could have any degree of complexity required for the mission, he adds. Even with its ROVs, Saab has always made efforts to provide a control system that simplifies things for the operator, by using onboard sensors to help stabilise the vehicle and make it easier and quicker to complete tasks. That also helps AUVs to operate safely and effectively when comms systems lack bandwidth and are characterised by significant latency and subject to dropouts. “Obviously you don’t want a vehicle that just keels over and floats away if something like that happens,” Siesjo says. “So what we have is a quite sophisticated step-by-step control scheme in which if you don’t specifically tell it to do something else it will just sit where it is. Then you can tell it to move forward 5 cm, say, move to the left or twist on its axis or whatever. “That allows you to do quite advanced things even over very bad comms links, because you can safely wait quite a while before you get feedback from the camera or other sensors that the vehicle has actually moved or that it is looking at the right object.” Behaviour-based architecture In the Sabertooth, Saab also implements a behaviour-based architecture for its autonomy, using algorithms that enable the vehicle to work towards more than one goal at a time. As Siesjo explains, “Imagine you tell the vehicle to go to a waypoint; if you then also tell it not to crash into something on the way then you basically have two potentially conflicting goals. What this type of algorithm does is try to manage those two goals by merging them. “Basically it says: ‘OK, I want to go there but I don’t want to get too close to that.’ So as the vehicle gets closer to the object it needs to avoid, it gets more and more negative input from the obstacle avoidance system, forcing it to deviate from the most direct track towards the waypoint and find a safer route. That is probably one of the simpler combinations.” Saab has provided obstacle avoidance capabilities on its vehicles for many years, typically relying on forward-looking sonars. Siesjo says early systems were extremely simple, in that when they detected an obstacle the vehicle would just stop until either the obstacle drifted out of its path or it received further instructions – a far cry from the modern system described above. The sonars on which the deep- sea AUV industry relies for collision avoidance these days are usually the same as those used for navigation, and Saab integrates units from several manufacturers including Imagenex, Norbit and Teledyne BlueView. The next level of control that Saab is looking at is the use of 3D simultaneous localisation and mapping in real time, which would allow the operator to lock the AUV onto a selected target with the sensors and have the vehicle guide itself there safely without the operator having to do much of anything. Siesjo adds that as AUV operators and their customers – the offshore energy companies – gradually get to grips with the degree of autonomy they want in their systems, they are looking to the categories used in the world of driver aids and autonomy in cars. “Here you have the simple stuff like cruise control, then following the road relatively well while still expecting the driver to take control immediately when prompted,” he says. “Ultimately you could reach the level of sophistication where you can tell the car to go home while you go to sleep in the back seat. That is where I think things are headed, but it is not part of what is specified for anything on these types of systems at the moment.” Open interfaces The hardware the control system runs on consists of industrial-standard computers running Windows Embedded software. Architecturally it is divided into two separate but connected systems: one that is open for customers to use more or less as Unmanned Systems Technology | February/March 2021 The Sabertooth during a demonstration at NASA’s Neutral Buoyancy Laboratory, where it communicated with the BlueComm optical comms system that offers data rates comparable with wi-fi (Courtesy of Saab)