Neural-network-based docking of autonomous vehicles

Abstract

In this paper, a neural-network-based guidance methodology is proposed for the docking of autonomous vehicles. The novelty of the overall system is its applicability to cases that do not allow for the direct proximity measurement of the vehicle's pose (position and orientation). In such instances, a guidance technique that utilizes line-of-sight based task-space sensory feedback is needed to minimize the impact of accumulated systematic motion errors. Herein, the proposed neural-network (NN) based guidance methodology is implemented during the final stage of the vehicle's motion (i.e., docking). The systematic motion errors of the vehicle are reduced iteratively by executing the corrective motion commands, generated by the NN, until the vehicle achieves its desired pose within random noise limits. The guidance methodology developed was successfully tested via simulations and experiments for a 3-dof high-precision planar platform