Issue 39 Unmanned Systems Technology August/September 2021 Maritime Robotics Mariner l Simulation tools focus l MRS MR-10 and MR-20 l UAVs insight l HFE International GenPod l Exotec Skypod l Autopilots focus l Aquaai Mazu

7 Platform one Unmanned Systems Technology | August/September 2021 Advanced Navigation has developed a range of inertial navigation systems (INSs) based around a fully digital fibre optic gyro (DFOG) architecture (writes Nick Flaherty). The Boreas D90 is the first INS to use the technology. The key is a new chip that integrates the optical interfaces, removing the need for optical splices to connect the chip to the coil of fibre that forms the heart of the gyroscope. The chip implements a specially developed digital modulation technique that passes spread-spectrum signals through the fibre optic coil and allows in-run variable errors in the coil to be measured and removed from the measurements. This makes a DFOG far more stable and reliable than traditional FOGs, by using a closed-loop accelerometer monitoring technique. This allows a smaller FOG with less coil length to achieve the accuracy of one with a longer coil. The DFOG was developed with Professor Arnan Mitchell, director of the Integrated Photonics and Applications Centre at RMIT University and an expert in single-chip optical integration. “By printing optical components onto a tiny chip, we are creating more compact and reliable fibre optic gyroscopes,” he said. The optical integration also provides a higher level of protection from shock, being able to resist up to 50 g for 11 ms, and vibration, resisting 8 g rms. Shock and vibration are two factors that traditionally reduce the accuracy of gyroscopes. The D90 combines the DFOG and accelerometer technologies with a dual- antenna RTK GNSS receiver that can provide centimetre-level accuracy when in range of an RTK network. These three elements are combined with a fusion algorithm that allows the INS to acquire a heading in less than 2 minutes when stationary or on the move, and maintains an accurate heading of 0.01 º secant latitude under all conditions with no reliance on GNSS. The dead-reckoning accuracy is 0.01% of the distance travelled measured using an odometer or Doppler velocity log.  That means the design of the fusion algorithm is critical to the performance of the INS. Rather than using a typical extended Kalman filter, the fusion algorithm is based on a machine learning technique that extracts far more information from the data by using a long/short-term memory (LSTM) neural network that is suited to classifying, processing and making predictions based on sensor data with a variable duration between important events. The LSTM algorithm developed by Advanced Navigation uses a combination of long-term learning that is hard-coded into the inference engine in the laboratory, while short-term learning operates in the field to update the model twice a second. This addresses deterministic errors from bias, scale-factor errors and non- orthogonal errors, as well as stochastic errors from sensor instabilities and noise. That means it can also be used for monitoring the health of the system. This LSTM algorithm gives the INS a bias stability of 0.001 º /hour that provides a roll/pitch accuracy of 0.005 º and a heading accuracy of 0.006 º . Both the hardware and software are designed and tested to IEC 61508 safety standards, and the unit has been environmentally tested to military standards. The system is designed for a mean time between failure of 500,000 hours. The power input is 9-36 V with 12 W power consumption with the GNSS. Additional connections include Ethernet, CAN and NMEA protocols, as well as a timing server providing Precision Time Protocol for clock synchronisation. The internal storage allows for up to a year’s worth of data logging. The D90 measures 160 x 140 x 115.5 mm, and weighs 2.5 kg, aiming the device at unmanned ground and marine applications. Navigation Fully digital FOGs unveiled The Boreas D90 provides centimetre-level accuracy when in range of an RTK network