Towards geometrization of control (Q1386879)

Professor Butkovskij is a man of great mathematical ability, and coauthor of perhaps the most important monograph ever written on control theory. In his latest publications [see for example, J. Comput. Syst. Sci. Int. 34, No. 5, 1-40 (1996); translation from Izv. Akad. Nauk, Teor. Sist. Upr. 1995, No. 4, 137-179 (1995; Zbl 0869.93011)] he offers deep discussions on the foundations of cybernetics, control theory, and occasionally of all physics. The present article is in part an exposition of unifying ideas of symmetry, calibration and invariants, that is, of conservation laws, in part a philosophical discussion on the foundations of cybernetics. In analogy with modern physics, which pursues unification of strong, weak, electromagnetic and gravitational fields, the author advocates a development of a Single (Unifying) Geometric Control Theory. The author comments that the limitations on the length of this exposition permits only an outline of his ideas. The introduction contains a philosophical discussion of similarities between modern control theory and modern physics. Cybernetics seems to overwhelm modern society by invading not only computer networks and technology, but also biology, medical fields, economics. The author offers a thorough discussion of commonly used concepts. Central mathematical notions are those of ``set'', ``mapping'', ``structure'', and (abstract) ``space''. The triad of the concepts of ``system'', ``structure'', ``information'' is essential to the synthesis of a control. Definitions given by Bourbaki and others are stated. For example, J. Dieudonné defines the structure assigned to a set \(S\) as a mapping \(C\) assigned to \(S\). More frequently, the offered informal definition is: ``structure is a collection of interactions between elements of a set''. While this is difficult to criticize, the basic flaw lies in the imprecise meaning of the term ``interaction''. The author prefers to regard \(C\) as relations assigned to various subsets of \(S\), to Cartesian products of \(S\), and unions of \(S\) with other sets. After discussing some types of structure that can be assigned, he offers a brief introduction to the possibility of \(C\) being a metric structure. Thus such a system may be regarded as an ordered pair \((S,C)\), that is, as a space (which he denotes by \(S_C)\). In this way he introduces the state of the system, both the local and the global state, as discussed in a previous article of the author. In his view, most areas of science could be regarded as legitimate subareas of control theory. For example, classical mechanics based on Newton's laws may be regarded as the study of the effects on the state \(x\), caused by the choices of the parameter \(F\), which we may recognize as a generalized force. Symmetry and symmetry-preserving transformations are the subject of the discussion in the next part of this paper. While any two real-life objects are always different in some way, in some cases they may be regarded as mathematically indistinguishable from each other. This implies that the metric assigned to some class of control problems makes it impossible to distinguish which is which. It is important that transformations defined as motions of the system (the author does give sufficient explanations) preserve the structure, and in particular cases that they preserve the metric. In more commonly used terminology, they preserve ``the geometry'', which generally has a more restricted interpretation. The author also notes that the concept of ``Control'' does not come from a generally accepted definition, and interpretations of its meaning may not resemble each other. Some scientists identify it with choice. Others associate it with the idea of variation. The idea of a Unifying Geometric Theory of Control (UFTG) is closer to the association of ``Control'' with ``Choice''. However ``Choice'' is best realized within mathematical set theory. Here the lack of choice may be realized by assignment of the empty set. Such choice may be made from a set of controllable parameters. Fixing the values of these parameters constitutes such a choice. The author ties elegantly these concepts to fibering and to the calibration of fields. Some ideas from a previous paper are explored in examples of control for systems with distributed parameters. He raises a number of important philosophical questions. Mathematical structures which are assumed to model physical systems appear to come from nowhere. One is reminded of a similar remark made by von Neumann. Most physicists stopped worrying about such problems, or about answers to questions like ``What is a vacuum?'', ``How do we interpret operators of creation and annihilation?'' The reviewer recalls an interesting article written about 20-25 years ago in Polish. The title was ``Properties of the empty set'', or similar. And it was far from trivial. The author pointed out that the apparent triviality of the empty set makes it difficult to unify set theory and analysis. Comment: The reviewer suggests that many such difficulties may be avoided by incorporating model theory of logic into basic concepts of both modern physics and control theory. In a suitable non-standard model certain properties will not vanish, but they shall cease to be observable. This may offer plausible mathematical tools to describe phenomena of ``annihilation'' and ``creation from a vacuum''. This article in a series of articles of the author contains important suggestions and comments. It deserves serious consideration by all control theoreticians. The English translation is seriously marred by bad translation errors. In the abstract: ``Единый'' is not uniform, it is ``single''. It could be stretched to ``unifying'', but should not be translated as ``uniform''. Собственные структуры об''ектов should read: ``fundamental structures of objects'', not ``eigenstructures of problems''. What is an eigenstructure of a problem? Page 13, line 9 ``matter of the problem'', would be better translated as ``nature of the problem''. Page 13 line 33 ``some unifying mappings'' should read ``multiple (or simultaneous) mappings'', line 34 ``by one mapping'' is awkward, ``by a single mapping'' sounds better. Page 13, second column line 14 ``two exemplars of the set \(S\)'', referring to the Cartesian product \(S\times S\), should be translated as two copies of \(S\). Line 31: Why is Russian translated into Latin? The translation ``ceteris paribus'' is basically correct, but why not stick to English: ``with all other conditions equal''? But let me skip minor flaws in this translation, such as ``centered parameters'', which should read ``lumped parameters'', and concentrate on a really serious mistranslation. Известный закон Фур'е should be translated as ``the well-known Fourier law'', not as ``the well-known Fourier transform''. It describes the Fourier law of heat conduction, or diffusion, and has nothing to do with the Fourier transform. A differential form is not ``precise'', it is exact, it does not become summable, it becomes integrable. We do not have groups of homologies and cohomologies, we have homology groups, cohomology groups, ect. My sympathy is with professor Butkovskij, whose article is written in excellent Russian, with deep knowledge of terminology used in a wide spectrum of mathematics, physics and engineering theories. The English translation does not reflect the high quality of his work.