Abstract:
The article considers the problem of controlling a linear system of differential equations under state constraints and pointwise restrictions on control parameters. Such problems are often found in applications from robotics, in the field of controlling autonomous movement on a plane or in space. As a rule, it is not possible to obtain an exact mathematical solution to such problems, and traditional numerical methods may be ineffective due to their slow speed. In recent years, methods for approximating solutions of control problems with state constraints using random graphs have become widespread. They have shown high efficiency in the case of trivial dynamics of a controlled object when it is possible to move without inertia along broken lines. However, problems with so-called kinodynamic constraints (when dynamics are described by nontrivial differential equations) have remained unsolved until recently. Significant progress in this area has been achieved by combining the ideas of constructing random graphs and ellipsoidal estimation techniques developed earlier by academician A. B. Kurzhanski and his students. This article continues the research in this area. The authors propose a new modification of the previously developed methods, which increases its effectiveness and makes it suitable for solving specific applied problems. The improvement of the method's characteristics was achieved by allocating the variables responsible for state constraints and separately processing such variables and the remaining part of the state vector.
Keywords:optimal control, state constraints, random graph, ellipsoidal calculus, ellipsoidal synthesis, reachability set.
Fulltext:
PDF file (2628 kB)
First page:
PDF file