Invited Session Proposal: New Developments in Closed-Loop Identification

2 Description of the topic Closed loop identification has always been recognized to be of primary importance and has been object of intense research since the late sixties. It is recognized that most of the times feedback cannot be removed for various reasons among which: (i) it may be intrinsic in the physical mechanism generating the data or (ii) data may come from an industrial plant where feedback loops cannot be open due to safety or production quality reasons or (iii) the physical mechanism might be too complex and may not be easily manipulated (think for instance to a communication network). Another reason for performing identification in closed-loop comes from the need to design experiments which guarantee a desired level of uncertainty in certain frequency bands. This has proven to be particularly useful in Identification for Control. The renewed interest for this subject in the recent years is due to a multiplicity of factors; on the " practical side " we mention the increased interest from various application areas (process control, control of networks and so on). Modern industrial applications often involve large plants described by tens of input and output variables and it is fair to say that these applications could only be tackled using the recently developed subspace methods (see e.g. contributions 5. and 6. below). On the other hand, there is a recent body of results concerning subspace methods in closed loop which have, for the first time, made these methods applicable to experiments in closed loop. There has recently been quite a lot of synergy between theory and applications. Nowadays researchers are well aware of the issues and difficulties arising when performing subspace identification using data collected in closed loop. The contributions of this session will touch upon several aspects of these, ranging from theory to practice. The contribution by K. Onodera, G. Emoto, and S. J. Qin deals with new algorithms for subspace identification (" A New Subspace Identification Method for Closed Loop Systems "); then S. J. Qin and L. Ljung will discuss the role of certain user choices in closed loop subspace identification (" On the Role of Future Horizon in Closed-Loop Subspace Identification "). The contribution by A. Chiuso and G. Picci (" Estimating the Asymptotic Variance of Closed Loop Subspace Estimators ") will try to bridge theory and practice showing that the recent theoretical results on the asymptotic properties of closed …