Radial Velocity and Bernouli Maximization Restricted Boltzmann Feedback Controller for Accurate Trajectory Tracking

Abstract

Accurate target tracking for rotary wing Unmanned Aerial Vehicle (UAV) is a fascinating application and a very demanding and complex field of research owing to the composite fluctuations and the diversified speed of moving target with respect to time. For this reason, several control algorithms have been evolved to track a target for rotary wing UAV. In this work a method called, Radial Velocity and Bernouli Maximization Restricted Boltzmann Feedback Controller (RV-BMRBFC) is introduced with the objective of suitable controller identification for accurate trajectory tracking in UAV. The RV-BMRBFC method is split into three sections, namely, data processing, controller identification and feedback controller for accurate trajectory tracking. First, the raw data obtained from Drone Dataset (UAV) is subjected to Partial Derivative Lagrangian-based Drone data processing for generating computationally efficient drone data for efficient controller identification. Second with the processed drone data as input Radial Velocity and Visual Axis Waypoint is applied for significant controller identification. The objective function in our work is formulated based on the response time, peak overshoot and settling time. By taking into consideration these objective function results, fitness is measured for all the processed controller identified results. Finally, Expected Bernoulli Maximization Restricted Boltzmann Machine-based Feedback Controller is applied with the identified controller positions for accurate trajectory tracking. With the obtained controller positions, target position data is said to be identified with which accurate trajectories are tracked in UAV. Experimental assessment is performed with diversified quantitative metrics like trajectory tracking accuracy, trajectory tracking time, trajectory tracking error rate and trajectory tracking overhead. The analyzed results demonstrate the superior performance of our proposed RV-BMRBFC method when compared with the two state-of-the-art methods.