Question
Asked 10 July 2015

Though position velocity and acceleration are dependent, why do we choose them as state variables?

It is preferred to choose state variables, which provide complete information of the system and must be linearly independent. are not position, velocity and acceleration related or in otherwords dependent? 

Most recent answer

Udaysimha Nerella
University of Duisburg-Essen
Thank you very much sir for your time and detailed explanation, now I got the point.  yes it would be great to discuss in person, but I am happy even though people like you are not reachable in person for newbies, platforms like these bring us on to one table.
Thank you 

All Answers (10)

Nikos Boretos
Mediterranean Agronomic Institute of Chania
Attaching a file, but essentially we want variables that allow an nth order DE to be reduce to n first order DEs which can be put in matrix form and are easier to solve...
See  http://www.roboticslab.ca/mass-spring-damper/ as well as Laplace transforms that replace the nth order DE with an nth order algebraic equation that again are easier to solve.
Udaysimha Nerella
University of Duisburg-Essen
Thanks for your answer Mr. Nikos Boretos, I do understand the concept of it, but my question is, as state variables need not be physically measurable or observable quanities , so anything related to the system can be taken as the state variable like heat of engine, brakes friction etc..., but to have complete controllability and observability of the system it is better to choose only the most important variable which gives the complete information of the system and they should be linearly independent else, we loose controllability or observability of the system (some valuable info is lost).
taking this into consideration choosing position and velocity as our state variables that is x and xdot , though they are normally have some relation as derivative of x is velocity, is this doesnot violate the linearly independency of x and xdot.
Nikos Boretos
Mediterranean Agronomic Institute of Chania
I believe linear independence applies to the variables as a whole, e.g. a,v and x, not just between a subset of the variables.
1 Recommendation
Udaysimha Nerella
University of Duisburg-Essen
then that would clear my doubt. Thanks for answering it.
Sir Soumitra Kumar Mallick
Institute for Advanced Study & IISWBM
Dear Prof. Nerella,
                                 In terms of the Copenhagen conception and the Heisenberg Principle these constants also act as variables as they are related amongst themselves by the uncertainty principle. So you have a point. But in terms of Schroedinger's wave conception of quantum mechanics and in terms of De Broglie number you need to consider these as state variables otherwise quantum calculations will become difficult if at all possible. I hope that helps. SKM QC
Udaysimha Nerella
University of Duisburg-Essen
Hello Mr. Soumitra K Mallick
At first I am very touched for giving me Prof. Title, because my goal is to be so, but unfortunately I am still a masters student but thanks for my future.
Coming to your answer, as you told x and xdot (position and velocity) have to be considered as state variables while considering in quantum levels.
When I use this state variable concept for mechanical electrical systems , I don't think we are considering quantum level Sir still aren't they dependent
Thanks and Regards
Udaysimha Nerella
Manfred Braun
University of Duisburg-Essen
The main point is: what does "dependence" mean? Of course, position, velocity and acceleration are related to each other by definition. But this is not an algebraic dependence. An algebraic dependence would be like "tell me your position and I shall tell you how fast you move". In this sense, position and velocity are independent. To evaluate the velocity v1 = v(t1) at a certain time t1 you need more than just x1 = x(t1), you need the position x = x(t) as a function of time in a whole interval, even a very small one, around the time instant t1. In the opposite way, you need the whole history v(t) up to the time instant t1 to evaluate the single value x1 = x(t1). 
If you had a facility that couples the velocity to the position in such a way that at any position x you have a unique velocity v = f(x), then you could omit v as an independent state variable, since it is uniquely determined by x, even if this function f is nonlinear and complicated. Usually there is no such coupling. Therefore x and v have to be kept as independent state variables. The relation xdot = v that holds between them is already part of the dynamical description of the system. 
2 Recommendations
Udaysimha Nerella
University of Duisburg-Essen
Hello Mr. Manfred Braun ,
Thankyou for your very detailed explanation, I think I understood it almost, there are few things I would like to get clarified on.
when we do this state space analysis in computer, to my knowledge  we use ode to get numerical solution so for velocity aren't we using a finite difference approximation, doen't this show algebraic relation between position and velocity?
also can you kindly explain me in more detail the following satement, "
To evaluate the velocity v1 = v(t1) at a certain time t1 you need more than just x1 = x(t1), you need the position x = x(t) as a function of time in a whole interval, even a very small one, around the time instant t1."
but the opposite way I understood it as we need to take integral over the time.
Thanks in advance
Manfred Braun
University of Duisburg-Essen
Of course, numerical simulation boils everything down to addition and multiplication of numbers, which is algebraic. But even then you don't get v at a certain time only from x at that time. Even in the simplest finite-difference approximation you need the positions at least for two neighboring time instances, and this is the discretized version of my previous statement that you need the function in an interval to get the derivative. So even in the discretized version there is no algebraic relation between x and v taken at the same time instant.
To get the derivative of x at a certain time instant t1, you need the function x(t) in a neighborhood of t1, otherwise you cannot form the derivative. From the single value x(t1) alone you don't get it. In the finite-difference approximation: you need at least x(t1) and x(t1 - h) to approximate xdot at t1.
If I were at our university, I would invite you to discuss it personally in my office. However, my answer comes from Tallinn, Estonia, where I am living.
2 Recommendations
Udaysimha Nerella
University of Duisburg-Essen
Thank you very much sir for your time and detailed explanation, now I got the point.  yes it would be great to discuss in person, but I am happy even though people like you are not reachable in person for newbies, platforms like these bring us on to one table.
Thank you 

Similar questions and discussions

Are the minions at Physics Stack Exchange Kuhnian? Please add to their discussion on string vibation theory.
Discussion
14 replies
  • Terence B AllenTerence B Allen
A minion is a low-level official protecting a bureaucracy form challengers.
A Kuhnian minion (after Thomas Kuhn's Structure of Scientific Revolutions) is a low-power scientist who dismisses any challenge to existing paradigm.
A paradigm is a truth structure that partitions scientific statement as true to the paradigm or false.
Recently, I posted a question on Physics Stack Exchange that serves as a summary of the elastic string paradigm. My question was: “Is it possible there can be a non-Fourier model of string vibration? Is there an exact solution?”
To explain, I asked if they knew the Hamiltonian equation for the string vibration. They did not agree it must exist. I pointed out there are problems with the elastic model of vibration with its two degrees of freedom and unsolvable equations of motion can only be approximated by numerical methods. I said elasticity makes superposition the 4th Newtonian law. How can a string vibrate in an infinite number of modes without violating energy conservation?
Here are some comments I got in response:
“What does string is not Fourier mean? – Qmechanic
“ ‘String modes cannot superimpose!’ Yet, empirically, they do.” – John Doty
“ A string has an infinite number of degrees of freedom, since it can be modeled as a continuous medium. If you manage to force only the first harmonic, the dynamics of the system only involve the first harmonic and it’s a standing wave: this solution does depend on time, being (time dependence in the amplitude of the sine). No 4th Newton’s law. I didn’t get the question about Hamilton equation.
“What do you mean with ‘archaic model’? Can I ask you what’s your background that makes you do this sentence? Physics, Math, Engineering? You postulate nothing here. You have continuum mechanics here. You have PDEs under the assumption of continuum only. You have exact solutions in simple problems, you have numerical methods approximating and solving exact equations. And trust me: this is how the branch of physics used in many engineering fields, from mechanical, to civil, to aerospace engineering.” – basics
I want to show the rigid versus elastic dichotomy goes back to the calculus wars. Quoting here from Euler and Modern Science, published by the Mathematical Association of America:
"We now turn to the most famous disagreement between Euler and d’Alembert … over the particular problem of the theory of elasticity concerning a string whose transverse vibrations are expressed through second-order partial differential equations of a hyperbolic type later called the wave equation. The problem had long been of interest to mathematicians. The first approach worthy of note was proposed by B. Taylor, … A decisive step forward was made by d’Alembert in … the differential equation for the vibrations, its general solution in the form of two “arbitrary functions” arrived at by means original with d’Alembert, and a method of determining these functions from any prescribed initial and boundary conditions.”
[Editorial Note: The boundary conditions were taken to be the string endpoints. The use of the word hyperbolic is, I believe, a clear reference to Taylor’s string. A string with constant curvature can only have one mathematic form, which is the cycloid, which is defined by the hyperbolic cosh x function. The cosh x function is the only class of solutions that are allowed if the string cannot elongate. The Taylor/Euler-d’Alembert dispute whether the string is trigonometric or hyperbolic.
Continuing the quote from Euler and Modern Science:
"The most crucial issue dividing d’Alembert and Euler in connection with the vibrating string problem was the compass of the class of functions admissible as solutions of the wave equation, and the boundary problems of mathematical physics generally, D’Alembert regarded it as essential that the admissible initial conditions obey stringent restrictions or, more explicitly, that the functions giving the initial shape and speed of the string should over the whole length of the string be representable by a single analytical expression … and furthermore be twice continuously differentiable (in our terminology). He considered the method invalid otherwise.
"However, Euler was of a different opinion … maintaining that for the purposes of physics it is essential to relax these restrictions: the class of admissible functions or, equivalently, curves should include any curve that one might imagine traced out by a “free motion of the hand”…Although in such cases the analytic method is inapplicable, Euler proposed a geometric construction for obtain the shape of the string at any instant. …
Bernoulli proposed finding a solution by the method of superimposition of simple trigonometric functions, i.e. using trigonometric series, or, as we would now say, Fourier series. Although Daniel Bernoulli’s idea was extremely fruitful—in other hands--, he proved unable to develop it further.
Another example is Euler's manifold of the musical key and pitch values as a torus. To be fair, Euler did not assert the torus but only drew a network show the Key and Pitch can move independently. This was before Mobius's classification theorem.
My point is it should be clear the musical key and pitch do not have different centers of harmonic motion. But in my experience, the minions will not allow Euler to be challenged by someone like me. Never mind Euler's theory of music was crackpot!

Related Publications

Article
The problem of estimation of the state-space model of a multivariable continuous-time system based on the observation of sampled input-output data is considered. The letter deals mainly with the determination of the continuous-time model from a linear discrete-time model estimated from the observations.
Article
In this part, a comparison between the different state-space models is presented. We discuss proper definitions of state, controllability and observability and their relations to minimality of 2-D systems. We also present new circuit realizations and 2-D digital filter hardware implementation of 2-D transfer functions.
Article
This paper presents an extension of the state-space approach to the study of discrete systems with finite delays of any magnitude relative to the discrete interval. The use of the state-vector equations for obtaining the response between the sampling instants is also demonstrated.
Got a technical question?
Get high-quality answers from experts.