Arithmetic constraints in Kalman filter implementation by using IMS A100 devices

Arithmetic constraints regarding digital Kalman filter implementation using digital signal processors with transversal filter architecture are outlined. Referring to the fixed gain Kalman filter, suitable results can be obtained as regards both computation speed and precision by using a particular state space representation. The proposed approach can represent a first step in implementing a high-order linear state-based regulator for fast dynamical system control. Studying the time varying gain Kalman filter realization by using the T212+IMS A100 hardware architecture, the 16-bit word length and the fixed point arithmetic due to IMS A100 devices lead to some drawbacks. However such difficulties are overcome by executing K(t) updating by using a concurrent algorithm running on the T414 transputer. >