Multivariable Adaptive Control of a Missile using MAC & CVA

Model Algorithmic Control (MAC) is a relatively new control technique which has been applied to many industrial processes during the last few years. Simplified analysis of the technique has shown that MAC, as an output based controller, has excellent robustness properties. It uses a model of the actual plant to compute an optimal control sequence by using an on-line optimization scheme. When the model is not known apriori, a relatively new technique based on Canonical Variate Analysis (CVA) can be used to identify the plant on-line. CVA, using Akaike Information Criteria (AIC), can determine the state-order and can identify a plant or a part of it in a closed loop configuration. It does not need any initialization and yet performs near maximum likelihood. In this paper, MAC and CVA has been integrated together to form an excellent robust adaptive controller known as AMAC (Adaptive MAC). The performance of AMAC has been demonstrated on an air-to-air missile model with asymmetric aerodynamic properties.