Unified Approach to Solving the Trajectory Problem by Usage of Neural Maps

Abstract

In modern robotics, there are many approaches to planning the trajectory of the portal manipulator. Methods based on the use of graphs, methods of potential fields, with the use of cellular decomposition, intelligent and others are used. Many varieties are based on similar mathematical apparatus. An important aspect of the trajectory search method applicability is convenience of chosen workspace and obstacles representation method. The approach proposed for manipulator path modelling implies a multi-level process, which includes periodic calculation and updating of invalid configurations set, considering the held cargo dimensions, a neural map creation and activation by the wave front propagation method etc. During experimental simulations with different activation functions, numbers of artificial neurons and synaptic connection implementation types, few methods of neural map activation have been considered, including flat and/or three-dimensional raster methods and wave propagation with static obstacles.