Development Of Airborne Intelligent Computing Unit For UAV Online Condition Monitoring

Unmanned aerial vehicle(UAV)are widely used in both military and civilian fields for their advantages of low cost and low casualties.However,due to the inability to obtain the flight status of UAVs in a timely manner,resulting in the crash of UAVs,the military and civilians suffer from different degrees of economic losses,so the timely acquisition of UAV flight status to ensure the safety of UAV flights has become a research hotspot.The existing complex state monitoring model is difficult to run under the constraint of limited computing resources for airborne deployment,which affects the accuracy and real-time of online monitoring.To address the above problems,this project will develop an airborne intelligent computing unit for online condition monitoring of UAVs.Through the real-time transmission of UAV operation status data combined with condition monitoring,intelligent model parallel calculation optimization,and multi-task scheduling optimization,this enables the accuracy and real-time requirements of UAV autonomous condition monitoring under a variety of airborne constraints.The details of the research are as follows:First,the overall design of the airborne intelligent computing unit for UAV online condition monitoring is proposed based on the functional and technical specifications.On this basis,the airborne data transmission interface module is designed to realize real-time transmission of UAV status data through fiber optic interface,and a high-performance core processing module is built based on Zynq Ultra Scale + MPSo C to meet the operational requirements of computational tasks such as parallel acceleration of status monitoring models and scheduling of multiple monitoring tasks in heterogeneous computing platforms.Next,the state monitoring unit is analyzed,and a state monitoring model based on a stacked long and short-term memory(LSTM)network,a custom computational acceleration unit based on a field-programmable gate array(FPGA),and a multi-task optimization model based on the particle swarm optimization algorithm(PSO)are designed.Among them,the stacked LSTM-based condition monitoring model completes the monitoring calculation of multiple flight parameters,and the FPGA-based custom computational acceleration unit completes the acceleration calculation of the monitoring model.The PSO-based multitasking optimization model completes the optimal deployment strategy of the custom computing acceleration unit and the multitasking scheduling optimization strategy under the dual constraints of computing resources and computing efficiency.At the same time,the fiber optic interface is designed to complete data transmission and data storage in fixed frame format,thus completing the construction of the state monitoring unit.Finally,the constructed airborne intelligent computing unit for online condition monitoring of the UAV is mounted on the semi-physical simulation platform,and then the technical indices of the airborne intelligent computing unit are verified and evaluated using the test data.The final test results show that the airborne intelligent computing unit meets the design requirements.