Aspects of Applying the Method of Coordinate Descent for the Shepherd Dog Bio-inspired Algorithm

Abstract

Further development of digital economy requires researching the limits of some algorithms applicability. These algorithms are based on the mechanisms obtained as a result of investigating the processes of “alive” nature. Such algorithms were called the bio-inspired algorithms and are widely used in practice, for example, for solving the problems of multi-agent systems coordinated control. Application of coordinate descent, a type of gradient method – extends the description of the existing shepherd dog algorithm. In the pasture, the actions of a sheep dog are a characteristic example when one agent forces many to move in the prescribed direction. Such approach is widely applicable in practice for crowd control, solving the problem of cleaning the environment and other engineering tasks. In spite of the fact that the heuristic algorithm of a shepherd dog is already described in scientific works, there is still a problem of defining the algorithm for the other agents’ movement. From the authors’ point of view, the selfish herd approach is formalized by the Gauss–Seidel method of coordinate descent, named similarly to the Gauss–Seidel iterative technique for solving a square system of linear equations.