lecture_5 - 2.160 System Identification, Estimation, and...

Info iconThis preview shows pages 1–4. Sign up to view the full content.

View Full Document Right Arrow Icon
2.160 System Identification, Estimation, and Learning Lecture Notes No. 5 February 22, 2006 4. Kalman Filtering 4.1 State Estimation Using Observers In discrete-time form a linear time-varying, deterministic, dynamical system is represented by x t + 1 x A t ) u B (1 += t t t nx 1 rx 1 where x t R is a n -dimensional state vector, u k R is an input vector, and A , B are t t matrices with proper dimensions. Outputs of the system are functions of the state vector and are represented with a -dimensional vector y R x 1 : t y = x H t (2) t t xn where H t R is an observation matrix . ), H t variables, one can simulate the system for predicting states and outputs in response to a time seq uence of inputs. S igure 4 his simulator may not workwell when ee F -1 below. T the model parameters are not ex actly k nown;actual outputs observed in the real system will differ from the predicted values. G iven those parameter matrices ( t , B A t and initial conditions of the state x t x + + u t B t A t x 0 H t y t t+1 unit time delay F igure 4 -1 Dynamic simulator of deterministic system 1
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
A dynamic state observer is a real-time simulator with a feedbackmechanism for recursively correcting its estimated state based on the actual outputs measured from the real physical system. S igure 4 e a standard feedbackcontrol ee F -2 below. Note that, unlik system, the discrepancy between the predicted outputs y ı t and the actual outputs y t from the real system are fed back to the model rather than the real physical system. Using a nx feedbackgain matrix L R , the state observer is given by t x A ı t u B t L ( y x ı t + 1 + + = y ı ) t t t t t (3) y ı = H t x ı t t T o differentiate the estimated state from the actual state of the physical system, the estimated state residing in the real-time simulator is denoted x ı t . With this feedback the state of the simulator will follow the actual state of the real system, and thereby estimate the state accurately. If the system is observable , convergence of the estimated state to the actual state can be guaranteed with a proper feedbackgain. In other words, a stable observer can forget its initial conditions;regardless of an initial estimated state x ı 0 , the observer can produce the correct state as it converges. T tate his is Luenberger’s S Observer for deterministic systems. + + u t B t A t 0 ı x H t y t l i l G L t _ + + x t t y ı t x ı 1 ı + t x Model unit time delay Rea Phys ca System ain F -2 Luenberger’s state observer for igure 4 deterministic linear system ı A special case of the above state observer is estimation of constant parameters θ . ee eq n S uation (17) in Chapter 2. Replacing the state transition matrix A t by the n x identity matrixand setting inputs to zero leads to a recursive parameter estimation formula in (2-17): 2
Background image of page 2
ı ( t ) = ı ( ı( θ ( t 1) + Κ ( t y ) t y ) ) (2-17) t T he difference from the previous parameter estimation problem is that in state estimation the state mak es “state transition” as designated by the state transition matrix A t and the input matrix B t
Background image of page 3

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Image of page 4
This is the end of the preview. Sign up to access the rest of the document.

This note was uploaded on 02/27/2012 for the course MECHANICAL 2.160 taught by Professor Harryasada during the Spring '06 term at MIT.

Page1 / 9

lecture_5 - 2.160 System Identification, Estimation, and...

This preview shows document pages 1 - 4. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online