FPGA Hardware Nonlinear Control Design for Modular Nanosatellite Attitude Control System

CubeSat attitude control systems must be compact, fast, and accurate to achieve success in space missions with stringent control requirements. Nonlinear control strategies allow the creation of robust algorithms for orbit and attitude control, but can have higher processing requirements in operation. In addition, It is desirable to implement distribute components of a CubeSat control system across hardware as much as possible to decrease the load on a central computing unit and to make the system more tolerant to point failures. Field Programmable Gate Array technology has become popular as a solution to the limited electronic space and more demanding speed requirements present in new-generation CubeSat systems. This paper will focus on demonstrating the feasibility and effectiveness of a proposed nonlinear adaptive fuzzy controller implemented on a Field Programmable Gate Array as part of a highly-integrated space hardware system that is under development. To facilitate integration with other system-on-a-chip CubeSat hardware, the new controller is implemented as a single logic block that can be added as a module to an FPGA-based system. Unlike the majority of FPGA-based controller systems that are based on soft-core processors, the controller is implemented directly in digital logic, saving resources on the FPGA fabric and increasing speed significantly. From a stream of sensor data, the logic-based adaptive fuzzy controller performs all filtering and calculations necessary for attitude control of the CubeSat, with the addition of a serial interface used for producing debugging logs. The difficulty of this approach is largely due to the limited functionality available in existing FPGA libraries. We overcome this by using freely-available High-Level Synthesis tools to convert proven C code into Hardware Description Language, which is then integrated on a modular level with the rest of the FPGAbased system. As full floating-point calculation requires high complexity in FPGA fabric, fixed-point arithmetic implementations are used in place of floating-point implementations as needed. A complete design of the resulting system is provided. Purely numerical simulations for the controller are done using Simulink on a host computer, and these simulations are compared with results obtained from hardware-in-the-loop testing of the FPGAbased controller using simulated sensor inputs. A practical comparison is also made between the high-precision floating point calculation of the simulations, and the more limited precision available on the FPGA hardware. Evaluation of the simulation and hardware test results shows that the control performance of the FPGA hardware control system is suitable for small satellite control, can meet precise pointing requirements, and that the algorithmic speed of the FPGAbased controller is significantly higher than the conventionallyprogrammed controller despite running at a much lower hardware clock speed. This implementation of the proposed controller shows the practicality of directly implementing nonlinear control algorithms directly on logic hardware as an alternative to microcontrollers and soft-core processors. Direct hardware implementations using modern synthesis methods are expected to continue to increase speed, reliability, and hardware modularity of CubeSat control algorithms by moving essential components into reconfigurable hardware. TABLE OF CONTENTS