Successive approximations of Bellman's function
Dynamic programming in optimal control and differential games (49L20) Numerical methods based on nonlinear programming (49M37) Numerical optimization and variational techniques (65K10) Dynamic programming (90C39) Nonlinear systems in control theory (93C10) Discrete-time control/observation systems (93C55)
The convergence of an algorithm of successive approximation of the Bellman function B for a discrete-time control problem \[ J=\sum^{K}_{k=1}f^ 0(k,x(k),u(k))+F(x(K))\to \min,\quad x(k+1)=f(k,x(k),u(k)),\quad x(0)=x^ 0, \] u(k)\(\in U(k,x(k))\), is proved. The j-th step of the algorithm: knowing the j-th approximation \(B_ j\) of B calculate \(u_ j(k,x)=\arg \min \{f(k,x,u)+B_ j(k+1,f(k,x,u)):\) \(u\in U(k,x)\}\), \(B_{j+1}(k,x)=J\) for \(u(i)=u_ j(i,x(i))\), \(i\geq k\), \(x(k)=x.\) The algorithm was considered by \textit{V. V. Kondrat'ev, V. P. Savel'ev} and \textit{O. S. Shorokhov} [Avtom. Telemekh. 1980, No.6, 48-57 (1980; Zbl 0485.93051)] for a narrower class of problems. Results of simulation on a computer are discussed. In particular, it is pointed out that \(u_ 1(k)=u_ 1(k,x(k))\) practically coincides with the precise solution of the traveling salesman problem with \(K=10\) and \(x\in R^ n\), \(n=10\).
- Extending Bell numbers for parsimonious chase estimation
- To the Bellman-Johnson problem
- scientific article; zbMATH DE number 3847201 (Why is no real title available?)
- scientific article; zbMATH DE number 3945922 (Why is no real title available?)
- Approximate algorithms for the traveling salesman problem. II
This page was built for publication: Successive approximations of Bellman's function
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1107099)
