hatimak/sigma

Sigma

FPGA implementation of sigma-point class of non-linear Gaussian filters, namely Unscented Kalman Filter (UKF), Cubature Kalman Filter (CKF), Cubature Quadrature Kalman Filter (CQKF) and Gauss Hermite Filter (GHF). Currently targeted towards Xilinx Zynq-7000 (ZC702) and Digilent Nexys4 DDR boards.

The intuition behind sigma-point class of filters is founded on the principle that it is easier to approximate a probability distribution than it is to approximate an arbitrary non-linear function or transformation. The members of this class of filters accomplish the approximation of a probability distribution by deterministically "sampling" support points (called sigma points) and assigning weight to each of these points. The mean and covariance can then be computed based on the statistics of these "sigma" points.

The implementation uses proprietary Xilinx IP for now and they come with their own set of copyright notices. The aim is to move away from proprietary IP and switch to open-source alternatives.

This work is done, under the supervision of Dr Shovan Bhaumik (Associate Professor, Dept of Electrical Engineering, Indian Institute of Technology Patna), for the partial fulfilment of requirements towards the award of a Bachelors degree. Please feel free to get in touch with the author, Hatim Kanchwala [email protected], for any details.

sigma's Issues

Create proprietary Xilinx IP on-the-go in Tcl scripts

Currently, the proprietary IP from Xilinx is generated locally, packaged and uploaded. The Tcl scripts then simply add the customised IP. Since this is a public repo it is more suitable to generate the IP with the required customisations via the Tcl scripts.

Report if input matrix is not positive definite

The Cholesky factorisation works only for positive definite matrices. The cholesky module should check for positive definiteness of the input matrix and report to the master.

It has been noted in simulations that this guard is essential because positive definiteness cannot be guaranteed throughout the iterations.

README needs to be descriptive

• Theoretical background
• Purpose
• Implementation details
• References