In this correspondence, we consider a partially decoupled variation of the
RLS algorithm. It is based on a constrained optimization of the cumulative
filter error using the higher order sets of filter weights to improve on th
e performance of the lower order weight sets whose values are already estab
lished. From this constrained optimization, a recursive algorithm is develo
ped whose form closely resembles the standard Volterra RLS algorithm hut wi
th structural differences that arise from eliminating the dependence of the
lower order weight sets on the higher order weight sets while retaining th
e dependence of the higher order weights on the lower order weights. The re
sulting algorithm, while suboptimal, requires less computational effort tha
n the fully coupled version, converges to steady state in the same amount o
f time, and is shown by example not to exhibit a substantial degradation in
performance.