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

86 diagnosing the vehicle’s overall condition. A third module, the data processing module, sends the data to the cloud and links to a V2X-based accident notification service that informs adjacent vehicles and infrastructures of the self-diagnosis result analysed by the OBD. A link to the cloud has a shorter transmission delay and ensures a higher transmission success rate. The timing impact introduced by various CAN/Ethernet multiplexing strategies at the gateway is used to derive upper bounds on end-to-end latencies for complex multiplexing strategies, which is the key for the design of safety-critical real-time systems. The various actual arriving orders of gateway messages are examined in detail. An explorative worst-case response time (WCRT) computation method reduced the WCRT by 24%, dramatically improving the responsiveness of the system. Autopilot monitoring Performance monitoring is increasingly integrated into the autopilot in a UAV, as it gathers all the sensor data. That means the performance monitoring can be handled through the autopilot via different basic alarms, such as failure in a comms port, the GNSS signal or an IMU failure. That can be taken further though. Telemetry analytics allow a UAV operator to load the log file from previous flights and analyse the data to graphically represent all the variables received through the telemetry, to see all the parameters and even make multiple comparisons of different data points. That allows performance monitoring services where users who have a problem or incident, or are unsure about the behaviour of certain parameters, to send the log file for analysis. The data also provides a graphical representation of the variables in real time, so that users can be performing the flight while seeing the telemetry from the autopilot. A replay mode uses the data to look back through the details of the flight to monitor the performance. That requires careful consideration of the data management. The configuration of the general-purpose I/O lines and flight paths are stored in non-volatile memory, where wear-levelling is an important factor in maintaining reliability. Diagnostic data is often recorded in the same page of a memory chip, then overwritten by the next set of data. That can mean certain pages are used more regularly, leading to a reduction in the reliability of storing the data as each read- write cycle wears down the memory cell, leading to the loss of data, one of the most common long-term failures in an electronic system. Wear-levelling algorithms move the data around a memory chip to record the data in different areas of it, ensuring even wear across the memory chip. The software also provides back-ups of data, so if the system detects an error it can automatically recover a copy from a redundant module of the memory and replace it. This is stored in a separate page on the same device to guarantee reliable operation, However, it means the data has to be stored with different ID tags and with tight controls on its location. Digital twins Another non-invasive way to monitor the performance of an autonomous system is to use a digital twin. This is a simulation of a complete vehicle or aircraft that is used for development and testing. A behavioural model of each component can be used to develop the overall system performance, then components can be replaced with detailed models of individual subsystems and chips. These models are used to develop the ML algorithms of platforms, driving or flying millions of virtual miles to enhance the algorithms. The models can also be used for monitoring the performance of the system after it has been developed. They can run 24 hours a day, seven days a August/September 2019 | Unmanned Systems Technology Performance monitoring is integrated into autopilot software to allow UAV operators to replay a flight with visibility of all the parameters (Courtesy of UAV Navigation) A QR code allows the condition of an individual battery pack to be uploaded to the cloud and combined with data from other packs in a fleet. This can also allow an individual pack to have a customised charging algorithm (Courtesy of Bosch)