Development Of On-Board Reconfigurable Computing System For UAV Trajectory Prediction

With the rapid development of aerospace technology, UAVs have been widely used in the field of military, civilian and consumption, because of its low cost and high flexibility. Trajectory prediction is one of key technologies to ensure the UAVs flight stable and safe, which sends the prediction results back to UAV Flight Control System and provides information support for system health management. Limited by requirements of real-time, power consumption and size on the UAV on-board platform, embedded reconfigurable computing system based on vector processor is chosen to provide universal computing platform for trajectory prediction implementation due to its low power and high performance fea tures.An on-board hardware platform is developed in this thesis with heterogeneous ZYNQ So C device as the core processor. Reconfigurable computing system is developed for the implementation of trajectory prediction based on the hardware platform and versatile, high-performance vector processor architecture. Moreover, case is designed by taken EKF algorithm for an example. Delay of data processing, performance and power of the system are tested and analyzed.Firstly, on-board reconfigurable computing system hardware platform is designed and developed. The hardware platform uses ZYNQ So C device which has the ability to control the serial tasks of general processor and FPGA parallel computing capabilities as core processor, and uses DDR3 SDRAM as data memory, uses asynchronous RS-422 interface for data exchange with UAV.Secondly, reconfigurable computing system is designed with vector processor based on hardware platform. The system is used to accelerate vector computing and to provide high performance computing platform, convenient peripheral communication interface and large capacity of external data storage for implementation of trajectory prediction. Vector processor and RS-422 communication controller are instantiated in ZYNQ PL part. Software program code is designed in PS part. The coordination design provides abilities of loading instructions dynamically and exchanging data with peripherals for reconfigurable computing system.Finally, EKF algorithm, which is the most typical algorithm of trajectory predictions, is taken for an example. The case of on-board reconfigurable computing system is designed and analyzed. The advantages of the system are stated according to testing results of storage capacity, delay of data processing, performance and power.