Navigational Planning By Constrained Hierarchical Neural Net

In this paper we propose a new scheme for the navigational p lanning of mobile robots by Artificial Neural Nets . The planner was realized by two cascaded neural nets, for which the first net filters sensory information by sat isfying a set of predefined constraints, and the second net generates control commands for the adjustment of velocity and direction of the robot . The proposed technique attempts to opt imize three criteria concurrent ly, such as (i) path traversal, (ii) t ime of traversal, and (iii) energy consumpt ion . This goal was accomplished on a mobi le robot Pioneer 2 D X by using three concurrent programs on client-server architecture. Program ' te leoperate ' helps to give the training instances to the constraint neural network, and program navigator helps to maintain the connect ion between the client and the server. Program 'mappe r ' is meant for the navigation on a specif ic path by using C H N N . The result is satisfactory, reaching the goal within a vicinity of a 5cm radius while avoiding all the static obstacles on the way to goal .