Application of a shrinking horizon model predictive control strategy is used to predict and control the end product quality of composite laminates produced by batch autoclave curing. The main contribution is the demonstration, using a laboratory-scale autoclave, of the feasibility and advantages of a control strategy that adjusts the batch recipe on-line to correct for unmeasured disturbances entering the process. Readily available, on-line secondary measurements are used in conjunction with the process model to predict (and hence control) quality-related end product properties. On-line monitoring is also used to monitor the process even-after the possibility of on-line correction has passed. The results show that this approach significantly reduces end Product quality variance.