Positioning error compensation of an industrial robot using neural networks and experimental study

TLDR

Industrial robots offer high efficiency and flexibility, yet their serial configuration limits positioning accuracy, restricting use in precision machining tasks such as aircraft assembly. This study proposes a neural‑network‑based method to enhance robot positioning accuracy. A neural network optimized with a genetic particle swarm algorithm predicts positioning errors, which are then used to compensate target points; the approach is validated through experiments on a KUKA KR 500‑3 robot in no‑load and drilling scenarios. Experiments show positioning errors drop from 1.529 mm to 0.344 mm and from 1.879 mm to 0.227 mm, yielding accuracy gains of 77.6 % and 87.9 % for the two conditions.

Abstract

Due to the characteristics of high efficiency, wide working range, and high flexibility, industrial robots are being increasingly used in the industries of automotive, machining, electrical and electronic, rubber and plastics, aerospace, food, etc. Whereas the low positioning accuracy, resulted from the serial configuration of industrial robots, has limited their further developments and applications in the field of high requirements for machining accuracy, e.g., aircraft assembly. In this paper, a neural-network-based approach is proposed to improve the robots' positioning accuracy. Firstly, the neural network, optimized by a genetic particle swarm algorithm, is constructed to model and predict the positioning errors of an industrial robot. Next, the predicted errors are utilized to realize the compensation of the target points at the robot's workspace. Finally, a series of experiments of the KUKA KR 500–3 industrial robot with no-load and drilling scenarios are implemented to validate the proposed method. The experimental results show that the positioning errors of the robot are reduced from 1.529 mm to 0.344 mm and from 1.879 mm to 0.227 mm for the no-load and drilling conditions, respectively, which means that the position accuracy of the robot is increased by 77.6% and 87.9% for the two experimental conditions, respectively.