Unmanned Systems Technology 027 l Hummingbird XRP l Gimbals l UAVs insight l AUVSI report part 2 l O’Neill Power Systems NorEaster l Kratos Defense ATMA l Performance Monitoring l Kongsberg Maritime Sounder

87 Performance monitoring | Focus week, in an accelerated environment, highlighting any problems as the platform ages. That can potentially give early warning of the failure of a particular component or subsystem that can then be checked in the real world. Having a digital model allows for detailed instrumentation, giving visibility of many elements such as processor or sensor cores in a platform that may not be accessible in the real-world version. Another key advantage of a digital twin is that a simulation model can be used to develop the training data for the multi-layered neural networks used in the performance monitoring subsystems. This training can be difficult in real-world systems as it could take many months of real-world testing to generate the models, and then many more to verify them. This process can be shortened markedly by using a digital twin to generate the training data. Battery monitoring A key area of autonomous vehicle monitoring is the battery system. This has traditionally been done via the battery management system (BMS), using a range of algorithms including Kalman to flag up changes in the pack’s performance. The average service life of modern lithium-ion batteries is eight to 10 years, or between 500 and 1000 charge cycles, and battery makers usually guarantee mileages of between 100,000 and 160,000 km. The use of faster and more powerful charging, up from 50 to 320 kW, and with higher numbers of charge cycles for an autonomous system, puts more pressure on battery packs. Stress, either from temperature or high current draw, makes cells age faster. The older the batteries get, the lower their performance and capacity, and the shorter the vehicle’s range. New cloud-based services are being developed that supplement individual vehicles’ BMSs. Algorithms in the cloud continually analyse the status of a battery and take appropriate action to prevent or slow down cell ageing. These measures can reduce the wear and tear on a battery by as much as 20%. Real-time data gathering is essential here, as the cloud service uses the data to optimise every recharging process and modify how the vehicle drives. This ‘battery in the cloud’ is being used by a Chinese autonomous vehicle provider. All the battery-relevant data, including ambient temperature and charging habits, is first transmitted in real time to the cloud, where ML algorithms evaluate the data. That provides visibility of the battery’s current status at all times, and can also give a reliable forecast of its remaining service life and performance. It also uses data from fleets of vehicles. This ‘swarm intelligence’ identifies more of the stress factors for vehicle batteries and identifies them more quickly. For example, fully charged batteries age more quickly at particularly high or low ambient temperatures, so the cloud service makes sure the batteries are not charged to 100% when conditions are too hot or too cold. Reducing battery charge by only a few percentage points protects against inadvertent wear and tear. As soon as a battery fault or defect is identified, the operator can be notified. This increases the chances that a battery can be repaired before it becomes damaged or stops working. The cloud service also optimises the recharging process itself. The algorithms in the cloud can calculate an individual charge curve for each recharging process, regardless of whether it takes place at home or elsewhere. That means the battery is recharged to the optimum level, helping to conserve the cells. Apps with charge timers monitor the recharging process so that it is carried out when demand for electricity is low. Optimising both fast and slow charging, and controlling the current and voltage levels during the charging process, helps to extend battery life. It also allows the cloud service provider to offer its own charging strategies, rather than those in the charger or BMS. That in turn can allow fast charging without damaging the battery, or a more leisurely standard charging process over several hours, enhancing the battery’s capacity and service life. Another driver for using a cloud service for performance monitoring is a lack of detailed technical knowledge about neural networks, especially in emerging markets. Sending the data to a service in the cloud for analysis means operators don’t need these specific skills but can still gain advantage from the latest performance monitoring technologies. A 23 Ah 48 V cell from LG Chem for example is being used in a portable battery that is compatible with common Unmanned Systems Technology | August/September 2019 A battery management system with a master and slave has been designed with a cloud connection to collect data from thousands of battery cells in vehicles to monitor their performance (Courtesy of Ion Energy)