1: \begin{abstract}

2: Finding the minimal bit rate needed for state estimation of a dynamical system is a fundamental problem in control theory. In this paper, we present a notion of topological entropy, to lower bound the bit rate needed to estimate the state of a nonlinear dynamical system with bounded inputs, up to a constant error $\eps$. We show that the usual notion of topological entropy that takes the limit as $\eps$ goes to zero and the one that requires exponential convergence of estimation error over time are not suitable for such systems. In particular, we show a simple system has an infinite entropy when using these notions. 

3: %

4: Since the actual value of our notion of entropy is hard to compute in general, we compute an upper bound instead. We show that as the bound on the input decreases, we recover a previously known bound on estimation entropy -- a similar notion of entropy -- for nonlinear systems without inputs \cite{LM015:HSCC}. For the sake of computing the bound, we present an algorithm that, given sampled and quantized measurements from a trajectory and an input signal up to a time bound $T>0$, constructs a function that approximates the trajectory up to an $\eps$ error up to time $T$. We show that this algorithm can also be used for state estimation if the input signal can indeed be sensed in addition to the state. Moreover, we present an improved bound on entropy for systems with linear inputs. 

5: %

6: We apply the computed upper bound to bound the entropy of nonlinear switched system and compare it with the bound in \cite{sibai-mitra-2017}. Finally, we show that an assumption made in \cite{sibai-mitra-2017} on the difference between the modes of the switched system is necessary by showing that a basic system that violates it has an infinite entropy. This paper is an extended version of our paper~\cite{sibai-mitra-2018}.

7: \end{abstract}

8: