State observers are very frequently used to estimate the states of the system to design controllers. The question is what are the disadvantages/limitations of using observers? Does output feedback control solve those problems? If not, then what are other methods to solve issues with observers?
Dear Pushkar Prakash Arya
This paper might be helpful, have a look:
Article Disturbance Observers: Methods and Applications. I. Methods
Kind Regards
Qamar Ul Islam
Dear Pushkar Prakash Arya,
The implementation of the control by state feedback u (x) needs sensors allowing to give at each instant ‘t’ an approximate value of the state x (t). Two types of sensors of different natures are used: The first is that of physical sensors, coming from the instrumentation. These sensors are sometimes too expensive or difficult to produce for technical reasons. For this reason, we have to design a second type of sensors, software sensors, more commonly called observers. These are algorithms based on a model of the system and using relevant information given by physical sensors. These software sensors provide an online estimate of unmeasured system state variables at any time.
The state is estimated by virtually copying the dynamics of the system, taking into account not only the command u, but also the outputs of the system (the measurements) y in order to correct any deviations.
For more details and information about this subject, i suggest you to see links on topic.
Article Observer design — A survey
https://link.springer.com/chapter/10.1007%2F978-0-8176-4661-5_9
https://intranet.ctism.ufsm.br/gsec/livros/Observers_in_Control_Systems.pdf
https://hal.archives-ouvertes.fr/tel-01465299/document
Best regards
The previous respondent was correct. Its not a real-time system because the observer reports the state after it is achieved. If you cant stand any latency in your system its a real problem. I would also add that the use of observers is based on a priori knowledge of the needed states and what each state "means" in terms of a controller response. I collect states and match them to a state table and when I find one I didnt consider beforehand I use both supervised and unsupervised learning techniques to add that state/response to my state table. This is only a problem when a large number of possible states exists. For a control system where you are just sensing "ON" and "OFF" this is not a problem at all. Good Luck.
Dear Pushkar Prakash Arya
Observers are a dynamic model of the real dynamic system, usually derived based on Newton's second-law of motion, which is itself an approximate model of the objective reality. Therefore, we cannot expect observers to fairly accurately estimate the states of the system. Since the states are based only on a hypothetical and approximate model of the objective reality, the control performance would be usually degraded without real sensors. But sensors are as well challenging due to measurement noise. This is why control realm still is not mature, and requires more research.
The main problem of the state observers is that the convergence and stablilty of closed-loop control system must be ansured for all operating space. However, in control of nonlinear systems these conditions are validated only locally or thay are dependent of some hypothesis which reduce its effectiveness in real applications.
An accurate model of the process is required for proper state estimation. Next, coefficients of the observer have to be chosen. This task can be accomplished using the pole placement technique. Despite the disadvantages mentioned above (i.e., phase shift, limited accuracy), observers are often used in electric drives.
Dear all, I am not an expert on control systems. But I wonder, the solution to your problems may not be metrology. In metrology, there are parameters to the differential equations describing the system. Sometimes accurate analysis of the system allows to significantly increase the accuracy and precision.Whether having such a solution of the problem you can further apply it to your problems.?
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Control Theory
- Control Systems