1: \documentclass[a4paper,11pt]{for_arXiv}
2: \usepackage{amsmath,amsfonts,amssymb}
3:
4: \renewcommand{\baselinestretch}{1.21}
5:
6: \newtheorem{example}[theorem]{Example}
7:
8: \newcommand{\al}{\mathbb{\alpha}}
9: \newcommand{\diam}{\diamondsuit_{\alpha}}
10: \newcommand{\A}{\mathcal{A}}
11: \newcommand{\C}{\mathcal{C}}
12: \newcommand{\D}{\mathcal{D}}
13: \newcommand{\dom}{{\rm dom}}
14: \newcommand{\emp}{\emptyset _0}
15: \newcommand{\f}{\varphi}
16: \newcommand{\F}{\mathcal{F}}
17: \newcommand{\h}{{\cal H}}
18: \newcommand{\id}{\mbox{\rm id}}
19: \newcommand{\K}{\mathcal{K}}
20: \newcommand{\La}{\Lambda}
21: \newcommand{\Lp}{\mathcal{L}}
22: \newcommand{\M}{\mathcal{M}}
23: \newcommand{\N}{\mathbb{N}}
24: \newcommand{\om}{\mbox{$\omega$}}
25: \newcommand{\ov}{\overline}
26: \newcommand{\R}{\mathbb{R}}
27: \newcommand{\rar}{\mbox{$\rightarrow$}}
28: \newcommand{\Si}{\Sigma}
29: \newcommand{\ti}{\tilde}
30: \newcommand{\T}{\mathbb{T}}
31: \newcommand{\tr}{\triangle}
32: \newcommand{\un}{\underline}
33: \newcommand{\U}{{\cal U}}
34: \newcommand{\Z}{\mathbb{Z}}
35: \newcommand{\Zi}{\mathcal{Z}}
36: \newcommand{\ha}{\mbox{}^{\al}\tilde{h}_k}
37: \newcommand{\hab}{\mbox{}^{\al}\tilde{h}_{k+1}}
38: \begin{document}
39:
40: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
41:
42: \title{Diamond-alpha Polynomial Series on Time
43: Scales\thanks{Accepted to the 8th Portuguese Conference
44: on Automatic Control -- CONTROLO'2008, 21 to 23 July 2008,
45: UTAD University, Vila Real, Portugal.}}
46:
47: \author{\textbf{Dorota Mozyrska$^1$ and Delfim F. M. Torres$^2$}}
48:
49: \date{$^1$Faculty of Computer Science \\ Department of Mathematics\\
50: Bia{\l}ystok Technical University\\
51: Wiejska 45A, Bia{\l}ystok 15-351, Poland\\
52: admoz@w.tkb.pl\\[0.3cm]
53: $^2$Department of Mathematics\\
54: University of Aveiro\\
55: 3810-193 Aveiro, Portugal\\
56: delfim@ua.pt}
57:
58: \maketitle
59:
60: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
61:
62: \begin{abstract}
63: \noindent \emph{The objective of this paper is twofold: (i) to survey existing
64: results of generalized polynomials on time scales, covering
65: definitions and properties for both delta and nabla derivatives;
66: (ii) to extend previous results by using the more general notion
67: of diamond-alpha derivative on time scales. We introduce a new
68: notion of combined-polynomial series on a time scale, as a convex
69: linear combination of delta and nabla generalized series. Main
70: results are formulated for homogenous time scales. As an example, we compute
71: diamond-alpha derivatives on time scales for delta and nabla
72: exponential functions.
73: }
74:
75: \medskip
76:
77: \noindent\textbf{Keywords:} time scales, diamond-$\alpha$ derivatives, generalized
78: polynomials, generalized series.
79:
80:
81: \medskip
82:
83: \noindent\textbf{2000 Mathematics Subject Classification:} 40C99, 39A13, 40A30.
84: \end{abstract}
85:
86: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
87:
88: \section{Introduction}
89:
90: Polynomial series are of great importance in control theory. Both
91: continuous and discrete polynomial series are useful in
92: approximating state and/or control variables, in modal reduction,
93: optimal control, and system identification, providing effective
94: and efficient computational methods \cite{Jacobsohn,Para,Sadek}.
95:
96: From recent years, the theory of control for discrete and
97: continuous time is being unified and extended by using the
98: formalism of time scales: see \cite{Zbig07,Zbig06} and references
99: therein. Looking to the literature on time scales, one understands
100: that such unification and extension is not unique. Two main
101: directions are being followed: one uses $\Delta$-derivatives while
102: the other chooses $\nabla$-derivatives instead. In this paper we
103: adopt the more general notion of diamond-$\alpha$ derivative
104: \cite{SFHD}, and give the first steps on a correspondent theory of
105: polynomial series. As particular cases, for $\alpha=1$ we get
106: $\Delta$-polynomial series; when $\alpha=0$ we obtain
107: $\nabla$-series. By choosing the time scale to be the real
108: (integer) numbers, we obtain the classical continuous (discrete)
109: polynomial series.
110:
111: Diamond-alpha derivatives have shown in computational experiments
112: to provide efficient and balanced approximation formulas, leading
113: to the design of more reliable numerical methods
114: \cite{Sheng1,SFHD}. We claim that the combined dynamic polynomial
115: series here introduced are useful in control applications.
116:
117: % --------------------------------------
118:
119: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
120:
121: \section{The $\Delta$, $\nabla$, and $\diam$ calculus}
122:
123: Here we give only very short introduction (with basic definitions)
124: on three types of calculus on time scales. For more information we
125: refer the reader to \cite{AB,And,Bh,Ozkan,Sheng,SFHD}.
126:
127: By a time scale, here denoted by $\T$, we mean a nonempty closed
128: subset of $\R$. As the theory of time scales give a way to unify
129: continuous and discrete analysis, the standard cases of time
130: scales are $\T=\R$, $\T=\Z$, or $\T=c\Z$, $c>0$.
131:
132: For $t\in\T$, the forward jump operator $\sigma$ and the
133: graininess function $\mu$ are defined by
134: $\sigma(t)=\inf\{s\in\T:s>t\}$ and $\sigma(\sup\T)=\sup\T$ if
135: $\sup\T<+\infty$; $\mu(t)=\sigma(t)-t$. Moreover, we have the backward
136: operator $\rho$ and the backward graininess function $\nu$ defined by $\rho(t)=\sup\{s\in\T: s<t\}$ and
137: $\rho(\inf\T)=\inf\T$ if $\inf\T>-\infty$; $\nu(t)=t-\rho(t)$.
138: In the continuous-time
139: case, \textrm{i.e.} when $\T=\R$, we have $\sigma (t)=\rho(t)=t$
140: and $\mu(t)=\nu(t)=0$ for all $t\in\R$. In the discrete-time case,
141: $\sigma(t)=t+1$, $\rho(t)=t-1$, and $\mu(t)=\nu(t)=1$ for each
142: $t\in\T=\Z$. For the composition between a function
143: $f:\T\rightarrow\R$ and functions $\sigma:\T\rightarrow\T$ and
144: $\rho:\T\rightarrow\T$, we use the abbreviations
145: $f^{\sigma}(t)=f(\sigma(t))$ and $f^{\rho}(t)=f(\rho(t))$.
146: A point $t$ is called
147: left-scattered (right-scattered) if $\rho(t)<t,$ ($\sigma(t)>t$). A
148: point $t$ is called left-dense (right-dense) if
149: $\rho(t)=t,$ ($\sigma(t)=t$).
150: The set
151: $\T^k$ is defined by $\T^k:=\T\backslash (\rho(\sup\T),\sup\T]$ if
152: $\sup\T<\infty$, and $\T^k=\T$ if $\sup\T=\infty$; the set $\T_k$
153: by $\T_k:=\T\backslash[\inf\T,\sigma(\inf\T))$ if $|\inf\T|
154: <\infty$, and $\T_k=\T$ if $\inf\T=-\infty$. Moreover, $\T^{k^{n+1}}:=\left(\T^{k^n}\right)^k,$
155: $\T_{k^{n+1}}:=\left(\T_{k^n}\right)_k$ and
156: $\T^k_k:=\T^k\cap \T_k$.
157:
158: For a function $f:\T\rightarrow\R$, we define the
159: $\Delta$-derivative of $f$ at $t\in\T^k$, denoted by
160: $f^{\Delta}(t)$, to be the number, if it exists, with the property
161: that for all $\varepsilon>0$, exists a neighborhood $U\subset \T$
162: of $t\in\T^k$ such that for all $s\in U$,
163: $|f^\sigma(t)-f(s)-f^{\Delta}(t)(\sigma(t)-s)|\leq
164: \varepsilon|\sigma(t)-s|$. Function $f$ is said to be
165: $\Delta$-differentiable on $\T^k$ provided $f^{\Delta}(t)$ exists
166: for all $t\in\T^k$.
167:
168: The $\nabla$-derivative of $f$, denoted by $f^{\nabla}(t)$, is
169: defined in a similar way: it is the number, if it exists, such
170: that for all $\varepsilon>0$ there is a neighborhood $V\subset \T$
171: of $t\in\T_k$ such that for all $s\in V,$
172: $|f^\rho(t)-f(s)-f^{\nabla}(t)(\rho(t)-s)|\leq
173: \varepsilon|\rho(t)-s|$. Function $f$ is said to be
174: $\nabla$-differentiable on $\T_k$ provided $f^{\nabla}(t)$ exists
175: for all $t\in\T_k$.
176:
177: \begin{example} The classical settings are obtained choosing $\T = \R$
178: and $\T = \Z$:
179: \begin{enumerate}
180: \item Let $\T=\R$. Then, $f^{\Delta}(t)=f^{\nabla}(t)=f'(t)$ and
181: $f$ is $\Delta$ and $\nabla$ differentiable if and only if it is
182: differentiable in the ordinary sense.
183:
184: \item Let $\T=c\Z$, $c>0$. Then,
185: $f^{\Delta}(t)=\frac{f(t+c)-f(t)}{c}$ and
186: $f^{\nabla}(t)=\frac{1}{c}\left(f(t)-f(t-c)\right)$ always exist.
187: \end{enumerate}
188: \end{example}
189:
190: It is possible to establish some relationships between $\Delta$
191: and $\nabla$ derivatives.
192:
193: \begin{theorem}{\rm \cite{Bh}}
194: (a) Assume that $f:\T\rightarrow\R$ is delta differentiable on
195: $\T^k$. Then, $f$ is nabla differentiable at $t$ and
196: $f^{\nabla}(t)=f^{\Delta}(\rho(t))$ for all $t\in\T_k$ such that
197: $\sigma(\rho(t))=t$. (b) Assume that $f:\T\rightarrow\R$ is nabla
198: differentiable on $\T_k$. Then, $f$ is delta differentiable at $t$
199: and $f^{\Delta}(t)=f^{\nabla}(\sigma(t))$ for all $t\in\T^k$ such
200: that $\rho(\sigma(t))=t$.
201: \end{theorem}
202:
203: A function $f:\T \rar \R$ is called rd-continuous provided it is
204: continuous at right-dense points in $\T$ and its left-sided limits
205: exist (finite) at left-dense points in $\T$. The class of real
206: rd-continuous functions defined on a time scale $\T$ is denoted by
207: $C_{rd}(\T,\R)$.
208:
209: If $f\in C_{rd}(\T,\R)$, then there exists a function $F(t)$ such
210: that $F^{\Delta}(t)=f(t)$. The delta-integral is defined by
211: $\int_{a}^bf(t)\Delta t=F(b)-F(a)$.
212:
213: Similarly, a function $f:\T \rar \R$ is called ld-continuous
214: provided it is continuous at left-dense points in $\T$ and its
215: right-sided limits exist (finite) at right-dense points in $\T$.
216: The class of real ld-continuous functions defined on a time scale
217: $\T$ is denoted by $C_{ld}(\T,\R)$. If $f\in C_{ld}(\T,\R)$, then
218: there exists a function $G(t)$ such that $G^{\nabla}(t)=f(t)$. In
219: this case we define $\int_{a}^bf(t)\nabla t=G(b)-G(a)$.
220:
221: \begin{example}
222: Let $\T=c\Z$, $c>0$, and $f\in C_{rd}(\T,\R) \cap C_{ld}(\T,\R)$.
223: Then, one has: $\int_{a}^bf(t)\Delta t=c\sum_{t=a}^{b-c}f(t)$,
224: $\int_{a}^bf(t)\nabla t=c\sum_{t=a+c}^{b}f(t)$.
225: \end{example}
226:
227: \begin{definition}\cite{Sheng1}
228: Let $\mu_{ts}=\sigma(t)-s$, $\eta_{ts}=\rho(t)-s$, and
229: $f:\T\rightarrow\R$. The diamond-alpha derivative of $f$ at $t$ is
230: defined to be the value $f^{\diam}(t)$, if it exists, such that
231: for all $\varepsilon>0$ there is a neighborhood $U\subset \T$ of
232: $t$ such that for all $s\in U,$
233: $$\left|\al\left[f^{\sigma}(t)-f(s)\right]\eta_{ts}
234: +(1-\al)\left[f^{\rho}(t)-f(s)\right]\mu_{ts}
235: -f^{\diam}(t)\mu_{ts}\eta_{ts}\right|\leq
236: \varepsilon|\mu_{ts}\eta_{ts}| \, . $$
237: We say that function $f$ is
238: $\diam$-differentiable on $\T^k_k$, provided $f^{\diam}(t)$ exists for all $t\in\T^k_k$.
239: \end{definition}
240:
241: \begin{theorem}\cite{Sheng1}\label{thsheng}
242: Let $f:\T\rightarrow\R$ be simultaneously $\Delta$ and $\nabla$
243: differentiable at $t\in\T^k_k$. Then, $f$ is $\diam$-differentiable at
244: $t$ and $f^{\diam}(t)=\al f^{\Delta}(t)+(1-\al)f^{\nabla}(t)$,
245: $\al\in[0,1]$.
246: \end{theorem}
247:
248: \begin{remark}
249: The $\diam$-derivative is a convex combination of delta and nabla
250: derivatives. It reduces to the $\Delta$-derivative for $\al=1$ and
251: to the $\nabla$-derivative for $\al=0$. The case $\al=0.5$ has
252: proved to be very useful in applications. For more on the theory
253: of $\diam$-derivatives than that we are able to provide here, we
254: refer the interested reader to \cite{Ozkan,Sheng1,Sheng,SFHD}.
255: \end{remark}
256:
257: The same idea used to define the combined derivative is taken to
258: define the combined integral.
259:
260: \begin{definition}\label{diaminteg}
261: Let $a,b\in\T$ and $f\in C_{rd}(\T,\R)\cap C_{ld}(\T,\R)$. Then,
262: the $\diam$-integral of $f$ is defined by
263: $\int_a^bf(\tau)\diam\tau=\al
264: \int_a^bf(\tau)\Delta\tau+(1-\al)\int_a^bf(\tau)\nabla\tau$, where
265: $\al\in[0,1]$.
266: \end{definition}
267:
268: In general the $\diam$-derivative of $\int_a^t f(\tau)\diam\tau$
269: with respect to $t$ is not equal to $f(t)$ \cite{Sheng}.
270:
271: Next proposition gives direct formulas for the $\diam$-derivative
272: of the exponential functions $e_p(\cdot,t_0)$ and
273: $\hat{e}_p(\cdot,t_0)$. For the definition of exponential and
274: trigonometric functions on time scales see, \textrm{e.g.},
275: \cite{Bh}.
276:
277: \begin{proposition}
278: \label{diamexpder} Let $\T$ be a time scale with the following properties: $\sigma(\rho(t))=t$, and $\rho(\sigma(t))=t$. Assume that $t_0\in\T$ and for all $t\in\T$ one has $1+\mu(t)p(t)\neq 0$, and $1-\nu(t)p(t)\neq 0$. Then,
279: \begin{gather}
280: e_p^{\diam}(t,t_0)=\left[\al p(t)+\frac{(1-\al)p^{\rho}(t)}{1+\nu(t)p^{\rho}(t)}\right]e_p(t,t_0) \, , \label{eq:a} \\
281: \hat{e}_p^{\diam}(t,t_0)=\left[(1-\al)p(t)+
282: \frac{\al p^{\sigma}(t)}{1-\mu(t)p^{\sigma}(t)}\right]\hat{e}_p(t,t_0) \, , \label{eq:b}
283: \end{gather}
284: for $t\in\T^k_k$.
285: \end{proposition}
286:
287: \begin{proof}
288: Firstly, recall that $f^{\sigma}= f+\mu f^{\Delta}$ and $f^{\rho}
289: = f-\nu f^{\nabla}$. Hence,
290: $e_p^{\nabla}(t,t_0)=p^{\rho}(t)e_p^{\rho}(t,t_0)
291: =p^{\rho}(t)\left(e_p(t,t_0)-\nu(t)e_p^{\nabla}(t,t_0)\right)$, and then
292: $e_p^{\nabla}(t,t_0)=\frac{p^{\rho}(t)}{1+p^{\rho}(t)\nu(t)}e_p(t,t_0)$, from
293: where it follows \eqref{eq:a}. Similarly, we have that
294: $\hat{e}_p^{\Delta}(t,t_0)=p^{\sigma}(t)\hat{e}_p^{\sigma}(t,t_0)
295: =p^{\sigma}(t)\left(\hat{e}_p(t,t_0)+\mu(t)\hat{e}_p^{\Delta}(t,t_0)\right)$,
296: and then
297: $\hat{e}_p^{\Delta}(t,t_0)=\frac{p^{\sigma}(t)}{1-p^{\sigma}(t)\mu(t)}\hat{e}_p(t,t_0)$,
298: from where \eqref{eq:b} holds.
299: \end{proof}
300:
301: \begin{corollary}
302: \label{cor1} Let $t,t_0\in\T$, $p(t)\equiv p$, and $1-\nu^2(t)p^2\neq 0$ for all $t\in\T$. Then, for $t\in\T^k_k$,
303:
304: (a) $\sin_p^{\diam}(t,t_0)
305: =\frac{p}{1+\nu^2p^2}((1+\al\nu^2p^2)\cos_p(t,t_0)$
306: $+(1-\al)\nu p\sin_p(t,t_0))$;
307:
308: (b) $\cos_p^{\diam}(t,t_0)
309: =\frac{-p}{1+\nu^2p^2}((1+\al\nu^2p^2)\sin_p(t,t_0)$
310: $-(1-\al)\nu p\cos_p(t,t_0))$;
311:
312: (c) $\sinh_p^{\diam}(t,t_0)
313: =\frac{p}{1-\nu^2p^2}((1-\al\nu^2p^2)\cosh_p(t,t_0)$
314: $-(1-\al)\nu p\sinh_p(t,t_0))$;
315:
316: (d) $\cosh_p^{\diam}(t,t_0)
317: =\frac{p}{1-\nu^2p^2}((1-\al\nu^2p^2)\sinh_p(t,t_0)$
318: $-(1-\al)\nu p\cosh_p(t,t_0))$.
319: \end{corollary}
320:
321: \begin{corollary}
322: Let $t_0\in\T$, $p(t)\equiv p$, and
323: $1-\mu^2(t)p^2\neq 0$ for all $t\in\T$. Then, for $t\in\T^k_k$,
324:
325: (a) $\widehat{\sin}_p^{\diam}(t,t_0)
326: =\frac{p}{1+\mu^2p^2}((1+(1-\al)\mu^2p^2)\widehat{\cos}_p(t,t_0)$
327: $ -\al\mu p\widehat{\sin}_p(t,t_0))$;
328:
329: (b) $\widehat{\cos}_p^{\diam}(t,t_0)
330: =\frac{-p}{1+\mu^2p^2}((1+(1-\al)\mu^2p^2)\widehat{\sin}_p(t,t_0)$
331: $ +\al\mu p\widehat{\cos}_p(t,t_0))$;
332:
333: (c) $\widehat{\sinh}_p^{\diam}(t,t_0)
334: =\frac{p}{1-\mu^2p^2}((1-(1-\al)\mu^2p^2)\widehat{\cosh}_p(t,t_0)$
335: $+\al\mu p\widehat{\sinh}_p(t,t_0))$;
336:
337: (d) $\widehat{\cosh}_p^{\diam}(t,t_0)
338: =\frac{p}{1-\mu^2p^2}((1-(1-\al)\mu^2p^2)\widehat{\sinh}_p(t,t_0)$
339: $ +\al\mu p\widehat{\cosh}_p(t,t_0))$.
340: \end{corollary}
341:
342: % --------------------------------------
343:
344: \section{Generalized monomials and polynomial series}
345:
346: Let $\T$ be an arbitrary time scale. Let us define recursively
347: functions $h_k:\T\times\T\rightarrow\R$, $k\in\N\cup\{0\}$, as
348: follows:
349: \begin{equation*}
350: h_0(t,t_0)\equiv 1 \, , \quad
351: h_{k+1}(t,t_0)=\int_{t_0}^th_k(\tau,t_0)\Delta \tau.
352: \end{equation*}
353:
354: Similarly, we consider the monomials $\hat{h}_k(\cdot,t_0)$: they
355: are the functions $\hat{h}_k:\T\times\T\rightarrow\R$,
356: $k\in\N\cup\{0\}$, defined recursively by
357: \begin{equation*}
358: \hat{h}_0(t,t_0)\equiv 1 \, , \quad
359: \hat{h}_{k+1}(t,t_0)=\int_{t_0}^t\hat{h}_k(\tau,t_0)\nabla \tau.
360: \end{equation*}
361: All functions $h_k$ are rd-continuous, all $\hat{h}_k$ are
362: ld-continuous. The derivatives of such functions show nice
363: properties: $h^{\Delta}_k(t,t_0) = h_{k-1}(t,t_0)$, $t\in\T^k$;
364: and $\hat{h}^{\nabla}_k(t,t_0)=\hat{h}_{k-1}(t,t_0)$, $t\in\T_k$,
365: where $t_0\in\T$ and derivatives are taken with respect to $t$. We
366: have that
367: \begin{equation*}
368: h_1(t,t_0)=\hat{h}_1(t,t_0)=t-t_0, \quad t, t_0\in\T \, .
369: \end{equation*}
370: For $\T=\R$, $h_k(t,t_0)=\hat{h}_k(t,t_0)=\frac{(t-t_0)^k}{k!}$.
371: Finding exact formulas of $h_k$ or $\hat{h}_k$ for an arbitrary
372: time scale is, however, not easy.
373: From \cite{HP} we have the following result:
374: \begin{theorem}
375: Let $t \in\T_k\cap\T^k$ and $t_0\in\T^{k^n}$. Then,
376: $\hat{h}_k(t,t_0)=(-1)^kh_k(t_0,t)$ for all $k\in\N\cup\{0\}$.
377: \end{theorem}
378:
379:
380: Next proposition gives explicit
381: formulas for homogenous time scales with $\mu(t)=c$, $c$ a
382: strictly positive constant. For that we need two notations of
383: factorial functions: for $k\geq 1$ we define $t^{\underline{k}} =
384: t(t-1)\cdots(t-k+1)$ and $t^{\overline{k}} = t(t+1)\cdots(t+k-1)$
385: with $t^{\underline{0}}:=1$ and $t^{\overline{0}}:=1$.
386:
387: \begin{proposition} Let $c>0$ and $\T=c\Z$.
388: For $k\in\N\cup\{0\}$ the following equalities hold:
389:
390: (a) $h_k(t,t_0)=c^k\binom{\frac{t-t_0}{c}}{k}
391: =c^k\frac{\left(\frac{t-t_0}{c}\right)^{\underline{k}}}{k!}$;
392:
393: (b) $\hat{h}_k(t,t_0)
394: =c^k\frac{\left(\frac{t-t_0}{c}\right)^{\overline{k}}}{k!}$.
395: \end{proposition}
396:
397: \begin{proof}
398: Firstly, $h_0(t,t_0)=\hat{h}_0(t,t_0)=1$. Next we observe that
399: $h_{k+1}^{\Delta}(t,t_0)=c^{k+1}\frac{\binom{\frac{\sigma(t)-t_0}{c}}{k+1}
400: -\binom{\frac{t-t_0}{c}}{k+1}}{c}\\
401: =c^{k}\left(
402: \binom{\frac{t-t_0}{c}+1}{k+1}-\binom{\frac{t-t_0}{c}}{k}\right)
403: =c^{k}\binom{\frac{t-t_0}{c}}{k}=h_{k}(t,t_0)$. Hence, by the
404: principle of mathematical induction, (a) holds for all
405: $k\in\N\cup\{0\}$. Since $\hat{h}_k(t,t_0)=(-1)^kh_k(t_0,t)$, (b)
406: is also true.
407: \end{proof}
408:
409:
410: \begin{remark}
411: Let $\T=c\Z$, $c> 0$. From the properties of factorial functions
412: it follows:
413: \begin{enumerate}
414: \item if $t\geq t_0\ \wedge \ k\geq \frac{t-t_0}{c}+1$, then
415: $h_k(t,t_0)=0$;
416:
417: \item if $t\leq t_0\ \wedge \ k \geq \frac{|t-t_0|}{c}+1$, then
418: $\hat{h}_k(t,t_0)=0$.
419: \end{enumerate}
420: In particular, when $c=1$ and $\T=\Z$, we have:
421: \begin{enumerate}
422: \item if $t\geq t_0\ \wedge \ k> t-t_0$, then $h_k(t,t_0)=0$;
423: \item if $t\leq t_0\ \wedge \ k> |t-t_0|$, then
424: $\hat{h}_k(t,t_0)=0$.
425: \end{enumerate}
426: \end{remark}
427:
428: In the next section we need the following results.
429:
430: \begin{remark}
431: Let $\T=c\Z$, $c>0$, and $t\in\T$. Then, for $k\in\N\cup\{0\}$, the following holds:
432:
433: (a) $\frac{t^{\underline{k+1}}}{t^{\underline{k}}} =t-k, \ t\leq k$;
434:
435: \bigskip
436:
437: (b) $\frac{t^{\overline{k+1}}}{t^{\overline{k}}}=t+k, \ t\geq -k$.
438: \end{remark}
439:
440: \begin{proposition}
441: \label{prop3012} Let $\T=c\Z$, $c>0$, $t,t_0\in\T$. Then,
442:
443: (a) $\lim\limits_{k\rightarrow
444: \infty}\left|\frac{h_{k+1}(t,t_0)}{h_k(t,t_0)}\right|=c$ \ \ for
445: $t<t_0$;
446:
447: (b) $\lim\limits_{k\rightarrow
448: \infty}\left|\frac{\hat{h}_{k+1}(t,t_0)}{\hat{h}_k(t,t_0)}\right|=c$\
449: \ for $t>t_0$.
450: \end{proposition}
451:
452: \begin{proof}
453: Let $\T=c\Z$, $c>0$, and $t,t_0\in\T$. For $t<t_0$, it is enough
454: to notice that
455: $\frac{h_{k+1}(t,t_0)}{h_k(t,t_0)}=\frac{t-t_0-ck}{k+1}$ to prove
456: (a). Equality (b) is proved in a similar way: for $t>t_0$, we have:
457: $\frac{\hat{h}_{k+1}(t,t_0)}{\hat{h}_k(t,_0)}=\frac{t-t_0+ck}{k+1}$.
458: \end{proof}
459:
460: \begin{remark}\cite{And}
461: Let $t_0\in\T^{k^n}$ and $k\in\N\cup\{0\}$. Then,
462: \begin{equation}\label{rel3}
463: \hat{h}_{k+1}^{\nabla}(t_0,t)=-\sum_{j=0}^k\nu^j(t)\hat{h}_{k-j}(t_0,t)
464: \end{equation}
465: for $\T_k\cap\T^k$.
466: \end{remark}
467: As a consequence of (\ref{rel3}) and equalities $f^{\sigma}= f+\mu
468: f^{\Delta}$ and $f^{\rho} = f-\nu f^{\nabla}$, the following laws
469: of differentiation of generalized monomials follow.
470:
471: \begin{corollary}
472: \label{derdor1}\mbox{}
473:
474: (a) $\left[h_{k+1}(t,t_0)\right]^{\nabla} =
475: \sum_{j=0}^{k}(-1)^j\nu^j(t)h_{k-j}(t,t_0)$;
476:
477: (b) $\left[\hat{h}_{k+1}(t,t_0)\right]^{\Delta}
478: =\sum_{j=0}^{k}\mu^j(t)\hat{h}_{k-j}(t,t_0)$.
479: \end{corollary}
480:
481: \begin{example}\label{exdor1}
482: Let $\T$ be an homogenous time scale with
483: $\mu(t)=\nu(t)=c=const.$, $c\geq 0$. Let us recall that for $c=0$
484: we have $\T=\R$ and for $c=1$ we have $\T=\Z$. Then,
485: $\left[h_{k+1}(t,t_0)\right]^{\nabla}
486: = \sum_{j=0}^{k}(-1)^jc^jh_{k-j}(t,t_0)$,
487: $\left[\hat{h}_{k+1}(t,t_0)\right]^{\Delta}=
488: \sum_{j=0}^{k}c^j\hat{h}_{k-j}(t,t_0)$.
489:
490: For $\T=\R$:
491: $\left[h_{k+1}(t,t_0)\right]^{\nabla}
492: =\left[\hat{h}_{k+1}(t,t_0)\right]^{\Delta} =\frac{(t-t_0)^k}{k!},$ \\
493: for $\T=\Z$:
494: $\left[h_{k+1}(t,t_0)\right]^{\nabla}=
495: \sum_{j=0}^{k}(-1)^jh_{k-j}(t,t_0)
496: =\binom{t-t_0}{0}-\binom{t-t_0}{1} +\cdots+(-1)^k\binom{t-t_0}{k},$
497: and $\left[\hat{h}_{k+1}(t,t_0)\right]^{\Delta}
498: =\sum_{j=0}^{k}\hat{h}_{k-j}(t,t_0)$.
499: \end{example}
500:
501: \begin{proposition} Let $t,t_0\in\T$. Then,
502:
503: (a) $h_1^{\diam}(t,t_0)=\hat{h}_1^{\diam}(t,t_0)\equiv 1$;
504:
505: (b) $h_{k+1}^{\diam}(t,t_0)=h_k(t,t_0)+(1-\al)\sum_{j=1}^k(-1)^j\nu^j(t)h_{k-j}(t,t_0)$;
506:
507: (c) $ \hat{h}_{k+1}^{\diam}(t,t_0)
508: =\hat{h}_k(t,t_0)+\al\sum_{j=1}^k\mu^j(t)\hat{h}_{k-j}(t,t_0)$.
509: \end{proposition}
510:
511: \begin{proof}
512: From the definition of $\diam$-derivative and
513: Corollary~\ref{derdor1}, we have:
514:
515: $h_{k+1}^{\diam}(t,t_0) =\al
516: h_{k+1}^{\Delta}(t,t_0)+(1-\al)h_{k+1}^{\nabla}(t,t_0) = \al
517: h_{k}(t,t_0)+(1-\al)\sum_{j=0}^{k}(-1)^j\nu^j(t)h_{k-j}(t,t_0)=
518: h_{k}(t,t_0)+(1-\al)\sum_{j=1}^{k}(-1)^j\nu^j(t)h_{k-j}(t,t_0)$.
519:
520: Next, $\hat{h}_{k+1}^{\diam}(t,t_0)=\al
521: \hat{h}_{k+1}^{\Delta}(t,t_0)
522: +(1-\al)\hat{h}_{k+1}^{\nabla}(t,t_0)=
523: \al
524: \sum_{j=0}^{k}\mu^j(t)\hat{h}_{k-j}(t,t_0)+(1-\al)\hat{h}_{k}(t,t_0)=
525: \hat{h}_{k}(t,t_0)+\al\sum_{j=1}^{k}\mu^j(t)\hat{h}_{k-j}(t,t_0)$.
526: \end{proof}
527:
528: \begin{theorem}\label{tayth1} {\rm \cite{HP}}
529: Assume that $f$ is $n+1$ delta-differentiable on $\T^{k^{n+1}}$.
530: Let $t_0\in\T^{k^{n}}$, $t\in\T$. Then,
531: \begin{equation}
532: \label{tay1}
533: f(t)= \sum_{k=0}^{n} f^{\Delta^k}(t_0)h_k(t,t_0)+R_n(t,t_0),
534: \end{equation}
535: where $R_n(t,t_0)=\int_{t_0}^{t}
536: f^{\tr^{n+1}}(\tau)h_{n}(t,\sigma(\tau))\Delta\tau$.
537: \end{theorem}
538:
539: \begin{theorem} \label{tayth2}{\rm \cite{And}} Assume that
540: $f$ is $n+1$ times nabla differentiable on $\T_{k^{n+1}}$. Let
541: $t_0\in\T_{k^{n}}$, $t\in\T$. Then,
542: \begin{equation}\label{tay2}
543: f(t)= \sum_{k=0}^{n}
544: f^{\nabla^k}(t_0)\hat{h}_k(t,t_0)
545: +\hat{R}_n(t,t_0),
546: \end{equation}
547: where $\hat{R}_n(t,t_0)=\int_{t_0}^{t}
548: f^{\nabla^{n+1}}(\tau)h_{n}(t,\rho(\tau))\nabla\tau$.
549: \end{theorem}
550:
551:
552:
553: By a polynomial real series we usually understand a series of the
554: form $\Sigma_{k=0}^{\infty} a_kP_k(t)$, where $(P_k(t))_{n \in
555: \N}$ is a given sequence of polynomials in the variable $t$ and
556: $(a_k)_{k \in \N}$ is a given sequence of real numbers. In the
557: continuous case one has $P_k(t)=\frac{(t-t_0)^k}{k!}$. For the
558: time scales we are considering in this paper, we have
559: $P_k(t)=h_k(t,t_0)$ or $\hat{P}_k(t)=\hat{h}_k(t,t_0)$, and we
560: speak about \emph{generalized power series on time scales}
561: \cite{BG,MP2,MP3}.
562:
563: \begin{definition}
564: \label{szereg} Let $\T$ be a time scale and let us fix $t_0\in\T$.
565: By a \emph{$\Delta$-polynomial series (on $\T$, originated at
566: $t_0$)} we shall mean the expression
567: $\sum_{k=0}^{\infty}a_kh_k(t,t_0)$, $t\in \T$; by a
568: \emph{$\nabla$-polynomial series (on $\T$, originated at $t_0$)}
569: we mean $\sum_{k=0}^{\infty}a_k\hat{h}_k(t,t_0)$, $t\in \T$, where
570: for each $k\in\N$, $a_k\in\R$. The sequence $(a_n)_{n\in\N}$ is
571: called the corresponding sequence of the series.
572: \end{definition}
573:
574: \begin{remark}
575: For any fixed $t, t_0\in\T$, both type of series become ordinary
576: number series. If they are convergent for $t$ we say that the
577: polynomial series is convergent at $t$.
578: \end{remark}
579:
580: If $\T=\Z$, then for each $t^*\in\Z$, $t^*\geq t_0$, the number
581: series $\sum_{k=0}^{\infty}a_kh_k(t^*,t_0)$ is convergent because
582: it is finite. The same situation we have when $t^*\leq t_0$:
583: $\sum_{k=0}^{\infty}a_k\hat{h}_k(t^*,t_0)$ is finite, so
584: convergent.
585:
586: In \cite{BG} and \cite{MP2} it is proved the following:
587: \begin{proposition}
588: Let $t_0\in\T$. If the power series
589: $\sum\limits_{k=0}^{\infty}a_k\frac{(t-t_0)^k}{k!}$ with the
590: corresponding sequence of coefficients $(a_k)_{k\in \N_0}$ is
591: convergent at $t^*\in\T$ and $t^*\geq t_0$, then the polynomial
592: series is convergent for all values of $t\in\T$ such that
593: $t_0<t<t^*\in\T$.
594: \end{proposition}
595:
596: Two polynomial series of the same type can be added
597: and multiplied by scalars giving the same type of series. We can
598: define the $\Delta$-derivative of $\Delta$-polynomial series:
599: $\left(\sum_{k=0}^{\infty}a_{k}h_k(t,t_0)\right)^{\Delta}=
600: \sum_{k=0}^{\infty}a_{k+1}h_k(t,t_0)$. Similarly, we have the
601: $\nabla$-derivative of $\nabla$-polynomial series in the form
602: $\left(\sum_{k=0}^{\infty}a_{k}\hat{h}_k(t,t_0)\right)^{\nabla}=
603: \sum_{k=0}^{\infty}a_{k+1}\hat{h}_k(t,t_0)$. Additionally, if the
604: $\Delta$-polynomial series is convergent for $t\in\langle
605: t_0,t^*)\cap\T$ and if the $\nabla$-polynomial series is
606: convergent for $t\in (t^*,t_0\rangle\cap\T$, then their
607: derivatives are also convergent on the same sets. From
608: Corollary~\ref{derdor1} we obtain the following result.
609:
610: \begin{proposition}
611: \label{derser} Let $ t_0\in\T$, $M>0$, and $\left(a_k\right)_{k\in\N\cup\{0\}}$ be a
612: sequence such that $|a_k|\leq M^k$ for each $k$. We have:
613:
614: (a) Let $I=\{t\in\T: \rho(t)\geq t_0 \ \wedge \ \nu(t)M<1\}$. Then, the series $\sum_{k=0}^{\infty}a_{k}h_k(t,t_0)$ is
615: convergent for $t\in I$, and
616: $\left(\sum_{k=0}^{\infty}a_{k}h_k(t,t_0)\right)^{\nabla}
617: =\sum_{k=0}^{\infty}\left(\sum_{j=0}^{\infty}
618: (-1)^j\nu^j(t)a_{j+k+1}\right)h_k(t,t_0)$
619: exists and it is convergent for $t\in I$.
620:
621: (b) Let $J=\{t\in\T: \sigma(t)\leq t_0 \ \wedge \ \mu(t)M<1\}$. Then, the series $\sum_{k=0}^{\infty}a_{k}\hat{h}_k(t,t_0)$ is
622: convergent for $t\in J$, and
623: $\left(\sum_{k=0}^{\infty}a_{k}\hat{h}_k(t,t_0)\right)^{\Delta}
624: =\sum_{k=0}^{\infty}\left(\sum_{j=0}^{\infty}
625: \mu^j(t)a_{j+k+1}\right)\hat{h}_k(t,t_0)$
626: exists and it is convergent for $t\in J$.
627: \end{proposition}
628:
629: \begin{remark}
630: Let $\T=c\Z$, $c>0$. There is no problem with convergence (i) in
631: points $t\geq t_0$ for series of the first type, (ii) at points
632: $t\leq t_0$ for series of the second (``hat'') type, because such
633: series are finite.
634: \end{remark}
635:
636:
637:
638: In \cite{BG} and \cite{MP2} one can find generalized series for an
639: exponential $e_p(t,t_0)$ with constant function $p(t)\equiv p$:
640: for $t\in\T$ and $t\geq t_0$ one has
641: $e_p(t,t_0)=\sum_{k=0}^{\infty}p^kh_k(t,t_0).$
642: It follows that $e_p^{\Delta}(t,t_0)
643: =p\sum_{k=0}^{\infty}p^kh_k(t,t_0)$, which gives the rule
644: $e_p^{\Delta}(t,t_0)=pe_p(t,t_0)$.
645:
646: As in Proposition~\ref{diamexpder}, let us consider now a time
647: scale with $\sigma(\rho(t))=t$. Then,
648: \begin{equation}
649: e_p^{\nabla}(t,t_0)
650: =\sum_{k=0}^{\infty}p^{k+1}\left(\sum_{j=0}^{\infty}
651: (-1)^j\nu^j(t)p^{j}\right)h_k(t,t_0) \, ,
652: \end{equation}
653: where
654: $\sum_{j=0}^{\infty}(-1)^j\nu^j(t)p^{j}=\frac{1}{1+p\nu(t)}$ if
655: $|p\nu(t)|<1$. This gives that
656: \begin{equation}
657: e_p^{\nabla}(t,t_0)=\frac{p}{1+p\nu(t)}\sum_{k=0}^{\infty}
658: p^kh_k(t,t_0)=\frac{p}{1+p\nu(t)}e_p(t,t_0)\, ,
659: \end{equation}
660: and then $e_p^{\diam}(t,t_0)$ is as in
661: Proposition~\ref{diamexpder}.
662:
663: In \cite{MP3} it is proved that
664: $\hat{e}_p(t,t_0)=\sum_{k=0}^{\infty}p^k\hat{h}_k(t,t_0)$ for
665: $t\geq t_0$. Then, we have that
666: $\hat{e}_p^{\nabla}(t,t_0)=p\hat{e}_p(t,t_0)$. We obtain that the
667: $\Delta$-derivative of $\hat{e}_p(\cdot, t_0)$ with respect to $t$ is given by the formula
668: $\hat{e}_p^{\Delta}(t,t_0)=\sum_{k=0}^{\infty}
669: p^{k+1}\left(\sum_{j=0}^{\infty}\mu^j(t)p^{j}\right)\hat{h}_k(t,t_0)$,
670: where $\sum_{j=0}^{\infty}\mu^j(t)p^{j}=\frac{1}{1-p\mu(t)}$ if
671: $|p\mu(t)|<1$. This gives that
672: $\hat{e}_p^{\Delta}(t,t_0)=\frac{p}{1-p\mu(t)}
673: \sum_{k=0}^{\infty}p^k\hat{h}_k(t,t_0)
674: =\frac{p}{1-p\mu(t)}\hat{e}_p(t,t_0)$
675: and then $\hat{e}_p^{\diam}(t,t_0)$ is also as in
676: Proposition~\ref{diamexpder}.
677:
678: % --------------------------------------
679:
680: \section{Combined series}
681:
682: The diamond-$\alpha$ derivative reduces to the standard $\Delta$
683: derivative for $\alpha=1$ and to the standard $\nabla$ derivative
684: for $\alpha=0$. The same ``weighted'' type definition is proposed
685: for the diamond-$\alpha$ integral. Based on this simple idea, we
686: introduce diamond type monomials.
687: Let us begin with the trivial remark that for any $f:\T\rightarrow
688: \R$ we can write $f(t)=\alpha f(t)+(1-\alpha)f(t)$.
689:
690: \begin{theorem}
691: Assume that $f$ is $n+1$ delta- and nabla-differentiable on
692: $\T^{k^{n+1}}$ and $\T_{k^{n+1}}$, respectively. Let
693: $t_0\in\T_{k^n}\cap\T^{k^{n}}$, $t\in\T$. Then,
694: \begin{equation*}
695: f(t)= \alpha S_n(t,t_0)+(1-\alpha)\hat{S}_n(t,t_0)+
696: \tilde{R}_n(t,t_0),
697: \end{equation*}
698: where
699: $S_n(t,t_0)=\sum_{k=0}^{n} f^{\Delta^k}(t_0)h_k(t,t_0) \, ,
700: \hat{S}_n(t,t_0) = \sum_{k=0}^{n}
701: f^{\nabla^k}(t_0)\hat{h}_k(t,t_0) \, $, and
702: $\tilde{R}_n(t,t_0)=\alpha R_n(t,t_0) + (1-\alpha)\hat{R}_n(t,t_0)
703: \, ,$
704: with remainders $R_n(t,t_0)$ and $\hat{R}_n(t,t_0)$ given as in
705: Theorems~\ref{tayth1} and \ref{tayth2}.
706: \end{theorem}
707:
708: \begin{definition}\label{diam_ser}
709: Let $\T$ be a time scale and $t_0\in\T$. By a
710: \emph{combined-polynomial series (on $\T$, originated at $t_0$)}
711: we shall mean the expression
712: \begin{equation}
713: \label{diam_ser1}
714: S_{\al}(t,t_0):=\al\sum_{k=0}^{\infty}a_kh_k(t,t_0)
715: +(1-\al)\sum_{k=0}^{\infty}b_k\hat{h}_k(t,t_0),
716: \end{equation}
717: where $t\in\T$ and $\al\in[0,1]$.
718: \end{definition}
719:
720: \begin{remark}
721: If in (\ref{diam_ser1}) we put $\al=1$, then we have a
722: $\Delta$-polynomial series. For $\al=0$ we obtain
723: $\nabla$-polynomial series. A combined-series is convergent if
724: both types of polynomial series are convergent. For fixed
725: $t,t_0\in\T$ we get usual number series, so we can say that the
726: series originated at $t_0$ is convergent at $t$ if it is
727: convergent as a number series.
728: \end{remark}
729:
730: \begin{proposition}
731: Let $\T=c\Z$, $c>0$, and $(a_0,a_1,\ldots)$, $(b_0,b_1,\ldots)$ be
732: two real sequences with nonzero elements such that
733: $\lim\limits_{k\rightarrow\infty}\left|\frac{a_{k+1}}{a_k}\right|
734: <\frac{1}{c}$,
735: $\lim\limits_{k\rightarrow\infty}\left|\frac{b_{k+1}}{b_k}\right|
736: <\frac{1}{c}$. Then, the combined-polynomial series
737: $$S_{\al}(t,t_0)=\al\sum_{k=0}^{\infty}a_kh_k(t,t_0)
738: +(1-\al)\sum_{k=0}^{\infty}b_k\hat{h}_k(t,t_0) \, ,$$
739: $\al\in[0,1]$, is convergent for all $t\in\T$.
740: \end{proposition}
741:
742: \begin{proof}
743: Based on Proposition \ref{prop3012}, we consider: $t=t_0$, when
744: combined-series $S_{\al}(t_0,t_0)=\al a_0+(1-\al)b_0$; $t>t_0$, when
745: the first part is finite the second is convergent; $t<t_0$, when
746: we have opposite situation to the previous one.
747: \end{proof}
748:
749: \begin{example}
750: Let $\T=\Z$ and $f(t)=2^t$. Then, $f^{\Delta^k}(t)=2^t$ and
751: $f^{\Delta^k}(0)=1$ for $k\in\N\cup\{0\}$. Additionally,
752: $2^t=\sum_{k=0}^{\infty}h_k(t,0)=\sum_{k=0}^t\binom{t}{k}$ for any
753: $t\geq 0$. But this series is not convergent for $t<0$. We have
754: $f^{\nabla^k}(t)=2^{t-k}$ and $f^{\nabla^k}(0)=2^{-k}$. The series
755: $\sum_{k=0}^{\infty}\frac{1}{2^k}\hat{h}_k(t,0)$ is convergent for
756: any $t\in\Z$. For that let
757: $c_k(t)=\frac{1}{2^k}\hat{h}_k(t,0)=\frac{t^{\bar{k}}}{2^kk!}$.
758: Then, $$\lim\limits_{k\rightarrow\infty}\frac{c_{k+1}(t)}{c_k(t)}
759: =\lim\limits_{k\rightarrow\infty}
760: \frac{(t+1)(t+2)\cdots(t+k)}{2^{k+1}(k+1)!}
761: \cdot\frac{2^kk!}{t(t+1)(t+2)\cdots(t+k-1)}
762: =\lim\limits_{k\rightarrow\infty}\frac{k+t}{2(k+1)t}=\frac{1}{2}$$
763: for each fixed $t>0$. The combined-polynomial series has the form $\al\sum_{k=0}^{\infty}\frac{t^{\underline{k}}}{k!}
764: +(1-\al)\sum_{k=0}^{\infty}\frac{1}{2^k}\frac{t^{\overline{k}}}{k!}$
765: and is convergent for $t\geq 0$.
766: \end{example}
767:
768: % --------------------------------------
769:
770: \section{Conclusions}
771:
772: Polynomial series have been used in the literature for solving a
773: variety of problems in control. In this paper we define Taylor
774: series via diamond-alpha derivatives on time scales and provide
775: the first steps on the correspondent theory. Such a theory
776: provides a general framework that is valid for discrete,
777: continuous or hybrid series. We trust that the polynomial series
778: here introduced are important in the analysis of control systems
779: on time scales.
780:
781: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
782:
783: \section*{Acknowledgments}
784:
785: The first author was supported by
786: Bia{\l}ystok Technical University grant S/WI/1/07;
787: the second author by the R\&D unit CEOC,
788: via FCT and the EC fund FEDER/POCI 2010.
789:
790: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
791:
792: \begin{thebibliography}{10}
793:
794: \bibitem{AB}
795: R.~P. Agarwal and M.~Bohner.
796: \newblock Basic calculus on time scales and some of its applications.
797: \newblock {\em Results Math.}, 35(1-2):3--22, 1999.
798:
799: \bibitem{And}
800: D.~R. Anderson.
801: \newblock Taylor polynomials for nabla dynamic equations on time scales.
802: \newblock {\em Panamer. Math. J.}, 12(4):17--27, 2002.
803:
804: \bibitem{Zbig07}
805: Z.~Bartosiewicz, {\"U}.~Kotta, E.~Paw{\l}uszewicz, and M.~Wyrwas.
806: \newblock Algebraic formalism of differential one-forms for nonlinear control
807: systems on time scales.
808: \newblock {\em Proc. Estonian Acad. Sci. Phys. Math.}, 56(3):264--282, 2007.
809:
810: \bibitem{Zbig06}
811: Z.~Bartosiewicz and E.~Paw{\l}uszewicz.
812: \newblock Realizations of linear control systems on time scales.
813: \newblock {\em Control Cybernet.}, 35(4):769--786, 2006.
814:
815: \bibitem{BG}
816: M.~Bohner and G.~S. Guseinov.
817: \newblock The convolution on time scales.
818: \newblock {\em Abstr. Appl. Anal.}, pages Art. ID 58373, 24, 2007.
819:
820: \bibitem{Bh}
821: M.~Bohner and A.~Peterson.
822: \newblock {\em Dynamic equations on time scales. An introduction with applications}.
823: \newblock Birkh\"auser Boston Inc., Boston, MA, 2001.
824:
825: \bibitem{HP}
826: R.~J. Higgins and A.~Peterson.
827: \newblock Cauchy functions and {T}aylor's formula for time scales {$\Bbb T$}.
828: \newblock In {\em Proceedings of the Sixth International Conference on
829: Difference Equations}, pages 299--308, Boca Raton, FL, 2004. CRC.
830:
831: \bibitem{Jacobsohn}
832: G.~Jacobsohn.
833: \newblock A discrete {T}aylor series method for the solution of two-point
834: boundary-value problems.
835: \newblock {\em J. Franklin Inst.}, 338(1):61--68, 2001.
836:
837: \bibitem{MP2}
838: D.~Mozyrska and E.~Paw{\l}uszewicz.
839: \newblock Functional series on time scales.
840: \newblock {\em Int. J. Math. Stat.}, 2(S08):94--105, 2008.
841:
842: \bibitem{MP3}
843: D.~Mozyrska and E.~Paw{\l}uszewicz.
844: \newblock Remarks on delta and nabla generalized polynomials and polynomial
845: series.
846: \newblock Technical report, Bialystok Technical University, 2008.
847:
848: \bibitem{Ozkan}
849: U.~M. Ozkan and H.~Yildirim.
850: \newblock Steffensen's integral inequality on time scales.
851: \newblock {\em J. Inequal. Appl.}, pages Art. ID 46524, 10, 2007.
852:
853: \bibitem{Para}
854: P.~N. Paraskevopoulos, A.~S. Tsirikos, and K.~G. Arvanitis.
855: \newblock New {T}aylor series approach to state-space analysis and optimal
856: control of linear systems.
857: \newblock {\em J. Optim. Theory Appl.}, 71(2):315--340, 1991.
858:
859: \bibitem{Sheng1}
860: J.~W. Rogers, Jr. and Q.~Sheng.
861: \newblock Notes on the diamond-{$\alpha$} dynamic derivative on time scales.
862: \newblock {\em J. Math. Anal. Appl.}, 326(1):228--241, 2007.
863:
864: \bibitem{Sadek}
865: I.~Sadek, T.~Abualrub, and M.~Abukhaled.
866: \newblock A computational method for solving optimal control of a system of
867: parallel beams using {L}egendre wavelets.
868: \newblock {\em Math. Comput. Modelling}, 45(9-10):1253--1264, 2007.
869:
870: \bibitem{Sheng}
871: Q.~Sheng.
872: \newblock Hybrid approximations via second order combined dynamic derivatives
873: on time scales.
874: \newblock {\em Electron. J. Qual. Theory Differ. Equ.}, pages No. 17, 13 pp.
875: (electronic), 2007.
876:
877: \bibitem{SFHD}
878: Q.~Sheng, M.~Fadag, J.~Henderson, and J.~M. Davis.
879: \newblock An exploration of combined dynamic derivatives on time scales and
880: their applications.
881: \newblock {\em Nonlinear Anal. Real World Appl.}, 7(3):395--413, 2006.
882:
883: \end{thebibliography}
884:
885:
886: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
887:
888: \end{document}
889: