AVM Navigator
AVM Navigator is an additional module of RoboRealm package (plugin) that provides object recognition and autonomous robot navigation with using of only one video camera on the robot as the main sensor of environment perception for navigation. It is possible due to using of algorithm "Associative Video Memory" (AVM) that is based on multilevel decomposition of recognition matrices and provides recognition of images with low False Acceptance Rate (AbOUT 0.01%). In this case the visual navigation for robot is just the sequence of images (landmarks) with associated coordinates that was memorized inside AVM tree during route training. The navigation map is presented as the set of data (such as X, Y coordinates and azimuth) that has associated with images inside AVM tree. And when robot sees the images from camera (marks) that can be recognized then it let defines his current location.
When target position is chosen on the map then navigator just CREATES way from current location to target position as the chain of waypoints. After this the navigator looks for nearest waypoint and goes to direction where point is. But if the real robot direction does not conform to direction of the current waypoint then navigator fixes it by turning of robot body to correct direction for further moving. When robot reaches current waypoint then navigator takes direction to other nearest waypoint in the chain and so further until the target position will be reached.