DENAVIT HARTENBERG NEWTON OPTIMIZED ITERATIVE DEMING REGRESSION BASED OPTIMAL TRAJECTORY TRACKING

Abstract

Unmanned Aerial Vehicles (UAVs) also referred to as drones are competent of performing mission related operations in an autonomous fashion. For the purpose of more accurately tracking designated trajectories with minimal response time and convergence speed, numerous trajectory tracking in UAV have been proposed. Progresses in computer and electronic technologies have smoothened evolution in automation control and intelligent algorithms and also certain significant contributions have been made into action on the trajectory tracking in UAV. In this work a method called, Denavit Hartenberg and Newton Optimized Iterative Deming Regression (DH-NOIDR) trajectory tracking in UAV with the objective of minimizing the response time and convergence speed is proposed. The DH-NOIDR method is divided into two parts, namely path planning and trajectory tracking. UAV path planning enables UAVs to key away from impediments and tracks the target in an efficient manner. To produce optimal paths without impediments collision for UAVs, a novel path planning algorithm based on 3D position information and frames of reference using Denavit Hartenberg parameters are initially proposed. With this type of design employing Denavit Hartenberg parameters results in the optimal path planning therefore reducing response time, and improving accuracy. Second, with the path planning results, a combination of TDoA and machine learning technique employing Deming Regression function is proposed that with the aid of three distinct characteristics, fine tuning the positioning of the drone, enhancing the iteration to ensure optimal search and setting the termination condition by reducing the sum of square residuals (SoSR) convergence-efficient trajectory tracking results are said to be obtained. The testing results have revealed that the DH-NOIDR method surpassed the state-of-the-art on the drone dataset in terms of response time, accuracy, error and convergence speed. In particular, the response time and convergence speed were reduced by 38% and 36% in comparison to earlier work respectively.