An FPGA-Based Unscented Kalman Filter for System-On-Chip Applications

The demand for fast and accurate state estimation in embedded systems has been increasing lately, due at least in part to mobile robotics such as an unmanned aerial vehicle (UAV). The desire to maintain high performance but with compact form factors leads to implementation issues, particularly with more complex systems. A hardware-based approach using field-programmable gate arrays may be able to alleviate these issues but tends to have a more complicated development process than traditional software-based approaches. In order to simplify development and promote portability between embedded applications, a hardware/software codesign of the unscented Kalman filter is presented. An example implementation (N = 18) of the hardware IP core only is presented using the Zynq-7000 XC7Z045 with synthesis, power, and timing results for the 1, 2, 5, and 10 processing-element cases.