Design method for an multidimensional neuronet based extrapolating path planner

We present a design method used to develop an extrapolating multidimensional neural network planner for mobile object with intelligent position-trajectory control system. The modeling results of a modified neural network neural network planner used for a robotic mobile object are presented. The design method is based on the bionic environment sensing in undefined conditions with stationary and mobile obstacles in a multidimensional space. The main design principal of the neural network planner structure is the hierarchical principle of information-processing systems. Based on this we present a hierarchical structure of a complex extrapolating multidimensional neural-alike network. Such network contains separate layers used on different stages to process the environment plan received from a robotic object's technical vision system. The hierarchical structure of a complex multidimensional neural-alike network is based on the following principles: object-oriented parametric synthesis, synthesis of weighted object position features with time sampling, and direction vector plans used to extrapolate such features. These plans with a certain probability determine the spatial position of related objects in future. We present the modelling results of selected methods used to detect round or spherical mobile obstacles based on technical vision system data and to predict their trajectories in two-dimensional and three-dimensional space. We present the results of software-based modelling of this approach to design neural network planner for the intelligent position-trajectory control system of mobile objects in two-dimensional space and in the simulation software package in three-dimensional space.