| A Controller Output Observer is used in estimating specific outputs that describe a physical system through the use of other outputs that are measured. Most of these signals can be measured directly. However, some signals are just easier to measure than others. This research is focused on a method to estimate immeasurable quantities, in real time, through the use of physical models and measured quantities. This can be done through the use of classical observers as introduced by Luenberger in 1964. However, since observers estimate the system states, one would have to again estimate the sought after quantities by another method, such as a constitutive relationship. The method proposed here shows that certain signals can be estimated directly by using a controller but without using a constitutive law. These estimated quantities are inputs to the model of the real system. In other words, these inputs drive the observer model. The method is tested using several vehicle models. Real measurements from a vehicle are used to show that the method succeeds in estimating quantities from a real vehicle. Another promising use for the Controller Output Observer is estimation in systems that include friction. This is because friction is very hard to model accurately. The Controller Output Observer does not require a model of the friction. The Controller Output Observer is also combined with the classic observer to use the advantage of both observers. The combined observer has the benefit of making the initial conditions decay at a prescribed speed. Possible applications for the estimated quantities include prognostics and health monitoring of systems, disturbance estimation and controls. An example of its use in vehicle controls is given. |
