D. S. Kurushin - Ph.D. (Eng.), Perm National Research Polytechnic University. Е-mail: daniel.kurushin@gmail.com
Е. V. Dolgova - Dr.Sc. (Econom.), Perm National Research Polytechnic University. Е-mail: elena.dolgova@gmail.com
R. А. Fayzrakhmanov - Dr.Sc. (Econom.), Professor, Head of Department, Perm National Research Polytechnic University. Е-mail: fayzrakhmanov@gmail.com

The article discusses the problem of controlling the movement of a mobile robot along an arbitrary terrain. The terrain simulated by a finite number of surfaces, and the position of the robot on a plane defined by (x, 0, y). The vertical coordinate is considered only from the perspective of the ability of a robot to cross the area. The computational complexity of two basic algorithms for the robot - object recognition and planning trajectory are estimated in the paper. Resolution of the images obtained from the vision system is 1392×1040 pixels, thus the matrix contains 1447680 binary elements for each of the two cameras. The map of robot's geoinformation system has the size 40×40 km in increments of 1 m. The number of objects in the ontology is 100, the number on the map - 100 000. The first algorithm consists of two main modules: an image recognition module and search for objects in the ontology. Search object in the ontology is equivalent to sorting of it's database, so complexity of this algorithm involves 100×10 = 1000 calls to the ontology. Object recognition depends on the assumed algorithm and the method of compressing of the original image. When calculating performance on the system which makes from 300 to 500 GFLOPs it can process about 10 images per cycle. For the second module estimated time routing cross-country on the basis of topographic maps is calculated. Through a series of transformations the image is represented in the form of a polygonal model. For each polygon based on the ontological domain model calculates the "color", characterizing the degree of passability of an area represented by this polygon. The results of a series of experiments on various maps are shown, depending on the obstacles computation time is from 5 to 30 s. Thus, in the paper solves the problem of mobile robot that is able to independently navigate it's way on a random terrain. The results will be used to create a software environment to control the movement of autonomous or semi-autonomous robot.