41 All this also has to be part of the functional safety assessment of the system design and integrated with recovery strategies in the event of a problem. Machine learning algorithms and artificial intelligence (AI) models are being used alongside sensors to predict problems so that repairs can be implemented before a failure occurs. However, there are also key differences in performance monitoring for different types of uncrewed systems. UAVs in the air, for example, see more monitoring of the tips of propellers as well as vibration monitoring of the motors, while UUVs might be many miles from assistance or completely inaccessible, making predictive monitoring vital. UAV monitoring Performance monitoring is key for safe and efficient UAV operations and it requires a combination of sensor data, the autopilot and ground control software. These work together to monitor hundreds of flight parameters in real time and trigger automatic safety actions and alerts for an operator. This requires multiple layers of hardware and logic redundancy and self-checking features, all within the size and weight constraints of the system. The redundancy includes duplicated components, such as microprocessor boards and altimeters, as well as sensor fusion algorithms to ensure that the estimation of performance is precise. This means that even if a position sensor fails or GNSS is lost, there is another alternative such as moving to dead reckoning. Layered on top of the hardware are continuous self-diagnostic routines that check the health of the sensors and other components in real time so that any anomaly is detected immediately to activate back-up systems. A powerful approach to performance monitoring is predefined autonomous actions. This requires a highly configurable autopilot to allow users to set automatic responses to strategic sensors such as setting different threshold values for sensors for particular actions. This could be in the event of the loss of communications or if vibrations are high. It means that an operator can define many different thresholds and customise the alarms, both in the UAV and in the ground station. However, multiple alarms can overwhelm a remote operator, so ‘light touch’ gauges can be used to visualise the alarms in different ways. These range from classic analogue and digital gauges to dynamic real-time graphics, allowing operators to switch views such as moving from looking at the battery voltage to observing something more intuitive such as mission time. Situational awareness A smart flag system can be used with multiple predefined thresholds. For example, for engine temperature, there can be a warning at 90 degrees and an alarm at 400 degrees, and rather than a standard beep, these can be Performance monitoring | Focus Machine learning algorithms and artificial intelligence models are being used alongside sensors to predict problems before a failure occurs Uncrewed Systems Technology | August/September 2025 Monitoring the performance of components in a UAV (Image courtesy of UAV Navigation-Grupo Oesía)
RkJQdWJsaXNoZXIy MjI2Mzk4