Reasoning about dynamical systems evolving over the reals is well-known to lead to undecidability. In particular, it is known that there cannot be reachability decision procedures for first-order theories over the reals extended with even very basic functions, or for logical theories that reason about real-valued functions, or decision procedures for state reachability. This mostly comes from the fact that reachability for dynamical systems over the reals is fundamentally undecidable, as Turing machines can be embedded into (even very simple) dynamical systems. However, various results in the literature have shown that decision procedures exist when restricting to robust systems, with a suitably-chosen notion of robustness. In particular, it has been established in the field of verification that if the state reachability is not sensitive to infinitesimal perturbations, then decision procedures for state reachability exist. In the context of logical theories over the reals, it has been established that decision procedures exist if we focus on properties not sensitive to arbitrarily small perturbations. For example by considering properties that are either true or δ-far from being true, for some δ > 0. In this article, we first propose a unified theory explaining in a uniform framework these statements, that were established in different contexts. More fundamentally, while all these statements are only about computability issues, we also consider complexity theory aspects. We prove that robustness to some precision is inherently related to the complexity of the decision procedure. When a system is robust, it makes sense to quantify at which level of perturbation it is. We prove that assuming robustness to a polynomial perturbation on precision leads to a characterisation of PSPACE. We prove that assuming robustness to polynomial perturbation on time or length leads to similar statements for PTIME. In other words, precision on computations is inherently related to space complexity, while length or time of trajectories, is intrinsically related to time complexity. These statements can also be interpreted in relation to several recent results about the computational power of analogue models of computation.

Models of computations over the integers are equivalent from a computability and complexity theory point of view by the (effective) Church-Turing thesis. It is not possible to unify discrete-time models over the reals. The situation is unclear but simpler for continuous-time models, as there is a unifying mathematical model, provided by ordinary differential equations (ODEs). Each model corresponds to a particular class of ODEs. For example, the General Purpose Analog Computer model of Claude Shannon, introduced as a mathematical model of analogue machines (Differential Analyzers), is known to correspond to polynomial ODEs. However, the question of a robust complexity theory for such models and its relations to classical (discrete) computation theory is an old problem. There was some recent significant progress: it has been proved that (classical) time complexity corresponds to the length of the involved curves, i.e. to the length of the solutions of the corresponding polynomial ODEs. The question of whether there is a simple and robust way to measure space complexity remains. We argue that space complexity corresponds to precision and conversely. Concretely, we propose and prove an algebraic characterisation of FPSPACE, using continuous ODEs. Recent papers proposed algebraic characterisations of polynomial-time and polynomial-space complexity classes over the reals, but with a discrete-time: those algebras rely on discrete ODE schemes. Here, we use classical (continuous) ODEs, with the classic definition of derivation and hence with the more natural context of continuous-time associated with ODEs. We characterise both the case of polynomial space functions over the integers and the reals. This is done by proving two inclusions. The first is obtained using some original polynomial space method for solving ODEs. For the other, we prove that Turing machines, with a proper representation of real numbers, can be simulated by continuous ODEs and not just discrete ODEs. A major consequence is that the associated space complexity is provably related to the numerical stability of involved schemas and the associated required precision. We obtain that a problem can be solved in polynomial space if and only if it can be simulated by some numerically stable ODE, using a polynomial precision.

In a recent article, the class of functions from the integers to the integers computable in polynomial time has been characterized using discrete ordinary differential equations (ODE), also known as finite differences. Doing so, we pointed out the fundamental role of linear (discrete) ODEs and classical ODE tools such as changes of variables to capture computability and complexity measures, or as a tool for programming. In this article, we extend the approach to a characterization of functions from the integers to the reals computable in polynomial time in the sense of computable analysis. In particular, we provide a characterization of such functions in terms of the smallest class of functions that contains some basic functions, and that is closed by composition, linear length ODEs, and a natural effective limit schema.