Grassmann inequalities and extremal varieties in \(\mathbb{P}\left(\bigwedge^p\mathbb{R}^n\right)\) (Q2046542)

This is a continuation of two articles co-written by one of the authors [the first author and \textit{G. Petroulakis}, Adv. Appl. Clifford Algebr. 27, No. 2, 1503--1515 (2017; Zbl 1378.51016); the first author et al., J. Optim. Theory Appl. 169, No. 1, 1--16 (2016; Zbl 1342.51010)]. The gist of their goal is a generalization of the study of extremal varieties of the Grassmannian of rank-\(p\) subspaces of \(\mathbb{R}^n\). Grassmannians are commonly amenable to real-life applications to engineering, control theory, signal processing and information theory. These applications are usually based on the characterization of varieties as locally ruled in a widely general sense, as well as intersection problems and embeddings of Grassmannian varieties into projective ones by way of the Plücker map. Part of these characterizations and problems are couched on the extremal variety problem, arising from a commonplace definition of a metric on projective spaces and the optimization of such distance from generic multivectors to a Grassmannian variety, in turn providing approximations of a multivector by a decomposable one, i.e. one expressible as \(v_1\wedge \dots \wedge v_p\). The \textit{extremal variety} is given by those vector approximations maximizing the distance, and its duality with respect to the original variety is of considerable interest from a theoretical standpoint. The aforementioned papers by one of the authors [the first author and Petroulakis, loc. cit.; the first author et al., loc. cit.] addressed this problem for \(p=2\) and the present paper addresses the more general case \(p>2\). The main computational tool is a least-squares approach to the \textit{Plücker relations}, which are the quadratic forms on the exterior power of \(V\) vanishing on the set of decomposable vectors. The resulting sum is linked to the Grassmannian matrix (representing the linear map \(T:\mathbb{R}^n \to \bigwedge^{p+1}\mathbb{R}^n\) defined by the wedge product with a fixed decomposable vector) via a wealth of old and new ancillary results as well as a very simple application of the Weierstrass extreme value theorem. The extremal variety \(\mathrm{Extr}(n,p)\) is eventually defined by a homogeneous quartic equation \(f_{n,p}(z)=M_{n,p}\| z\|^4\) in the projective Plücker space, depending on certain coefficients \(M_{n,p}\) whose computation becomes one of the relevant points of this paper. To be precise, these numbers are provided with a simple upper bound in general, as well as explicitly computed for specific instances of \(p\) and \(n\). These instances are not all-encompassing but they do offer considerable generality in their scope. The study of these numbers \(M_{n,p}\) and the general structure of the variety \(\mathrm{Extr}(n,p)\) is made considerably simpler by the reduction of a problem arising from multilinear algebra to the realm of linear algebra and normed spaces, once the Grassmann matrix makes its appearance, and this leads directly to a general bound on \(M_{n,p}\). The equations defining \(\mathrm{Extr}(n,p)\) adopt simplified forms for particular values or sets of values of \(n\) and \(p\), thereby producing the explicit values mentioned in the above paragraph. The last section deals with applications to state-of-the-art problems in control theory, namely pole placement for specific multiple-input and multiple-output systems. A very short and simple example related to the tangent bundle of a torus is included at the end of the article.