A new optimization-based iterative learning control algorithm is propo
sed and its properties derived. An important characteristic of this al
gorithm is that it uses present and future predicted errors to compute
the current control, in a similar manner to model-based predictive co
ntrol using a receding horizon. In particular, it enables the algorith
m designer to achieve good control over convergence rate. The actual i
mplementation has a multimodel structure but uses standard linear quad
ratic regulator methods for a causal formulation (in the iterative lea
rning sense) of what is originally a non-causal algorithm. The results
are illustrated by simulations.