Intelligent Sampling Technique for Path Tracking Controllers

Abstract

Path tracking controllers in mobile robots are aimed at driving the vehicle following a previously defined path. For implementation purposes, the path is discretized into samples. Each sample is theoretically reachable by the robot. Nevertheless, the motion commands could exceed the actuators’ limit causing their corresponding saturation, especially when maneuvering in constrained spaces. In this brief, a technique is proposed and implemented for an intelligent sampling of the path to be tracked by an automated vehicle. On the one hand, it is based on finding the best reference in a path in order to improve the controller’s performance without compromising the kinematic restrictions of the robot. On the other hand, it uses a probabilistic framework to predict and to avoid collision with moving objects. The proposed methodology is implemented in real time and tested on an agricultural machinery while monitoring a grove, sharing the workspace with field workers, and using the previously published path tracking controllers. The results obtained showed an improvement of 17% (average) of the controllers’ performance and an improvement of 40% (average) in the time associated with the navigation process.