Unmanned Systems Technology 021 | Robot Aviation FX450 l Imaging Sensors focus l UAVs Insight l Liquid-Piston X-Mini l Riptide l Eurosatory 2018 show report l Zipline l Electric Motors focus l ASTS show report

97 “So we’re developing software to automate the navigation, and we do it progressively, by starting with situational awareness sensors and other sorts of systems that help the onboard crew make more informed decisions.” The initial process typically involves integrating optical sensors to replace ‘lookout’ officers who can become tired or distracted at important moments. Cameras are selected for their trade- off between accuracy, redundancy and cost, as in autonomous car engineering. After that, the company’s software fuses the pertinent sensor inputs to produce actionable information outputs for the control officers. Work on path planning and course recommendation follows from automating situational awareness. “And once the system knows where the crew and cargo want to navigate, the decision training we conduct focuses on whether to tell the actuators to operate the steering and gas to follow the safest trajectory, or to minimise fuel consumption, or take any other mission or mechanical constraints into account,” Vollmer added. Also presenting marine autopilot systems was Sea Machines Robotics. Its SM300 is designed to provide ‘man- in-the-loop’ (Level 3) autonomy with collision avoidance capabilities provided by radar and AIS. As Peter Holm told us, “Users can also perform mission planning, using pre- set options such as setting up a survey mission, route points for following and collaborative operations. “The system can also integrate with any pre-existing onboard controls, whether they are fully electric or mechanical, such as cable-controlled rudders and hydraulic systems.” The collaborative following function can be programmed to set a fixed distance between target vessels using GPS, or it can be upgraded to use data from other sensors for a more accurate and/or dynamic configuration. The SM300 consists of an onboard control hub housed in an IP67-rated aluminium enclosure, and a remote user interface, which gives the option of using a ruggedised laptop or industrial-grade remote controller. Sea Machines also offers the SM400 situational awareness system that features AI-based target identification. “Training the AI in identification is the main challenge with systems like these,” Holm said. “And you need to train it using not only cameras but also Lidar, along with other identification methods as well.” “Our team in Boston is using an artificial neural network to ensure the sensors always know with the highest possible degree of confidence what objects are surrounding a ship.” Sonardyne showcased its range of sensors used in phase one of the AutoMinder (Autonomous Marine Navigation in Denied Environments) project. The Innovate-UK funded initiative is aimed at integrating modern sensor types into marine positioning systems in GNSS- denied areas, and Sonardyne partnered with Guidance Marine on the project. As Geraint West explained, “Offshore drill ships can cost hundreds of thousands of pounds a day to run, so any downtime can be extremely expensive; fully redundant alternatives to GNSS-based navigation are therefore critical requirements. “Dynamic positioning typically takes multiple different sensor solutions such as radar, Lidar and vision, and feeds them into a dynamic model – a Kalman filter – to inform our navigation state for controlling the vessel.” However, all sensors are prone to certain failure modes, for various environmental, physical or obstacle- related reasons. Using a combination of inertial navigation system (INS) and acoustic marine sensors such as doppler velocity logs (DVLs) and other sensors can constrain INS drift and ensure safe navigation of an autonomous vessel independently of GNSS-based navigation. The company has used its Sprint-Nav 500 INS, which integrates ring laser gyros, accelerometers and a Syrinx DVL, to output precise positioning accurate to less than 0.04% of distance travelled, while typically consuming 15 W of power. West added, “Our work with AUVs and ROVs has informed much of our approach to GNSS-denied navigation, because underwater there is no GNSS. The INS has an inbuilt ability to compensate for the distance between the DVL’s transducer and the angular rate, and acoustic sensors aid in constraining INS drift.” Unmanned Systems Technology | August/September 2018 Sea Machines Robotics’ SM300 autopilot provides ‘man-in-the-loop’ autonomy