A survey Of learning-Based control of robotic visual servoing systems

• Published in: Journal of the Franklin Institute Vol. 359; no. 1; pp. 556 - 577
• Main Authors: Wu, Jinhui, Jin, Zhehao, Liu, Andong, Yu, Li, Yang, Fuwen
• Format: Journal Article
• Language: English
• Published: Elmsford Elsevier Ltd 01.01.2022; Elsevier Science Ltd
• Subjects: Algorithms; Control systems; Decision making; Learning; Machine learning; Mathematical models; Neural networks; Predictive control; Robot control; Robotics; Robots; Servocontrol; Task complexity; Vision; Visual control
• ISSN: 0016-0032; 1879-2693; 0016-0032
• DOI: 10.1016/j.jfranklin.2021.11.009

Major difficulties and challenges of modern robotics systems focus on how to give robots self-learning and self-decision-making ability. Visual servoing control strategy is an important strategy of robotic systems to perceive the environment via the vision. The vision can guide new robotic systems to complete more complicated tasks in complex working environments. This survey aims at describing the state-of-the-art learning-based algorithms, especially those algorithms that combine with model predictive control (MPC) used in visual servoing systems, and providing some pioneering and advanced references with several numerical simulations. The general modeling methods of visual servo and the influence of traditional control strategies on robotic visual servoing systems are introduced. The advantages of introducing neural-network-based algorithms and reinforcement-learning-based algorithms into the systems are discussed. Finally, according to the existing research progress and references, the future directions of robotic visual servoing systems are summarized and prospected.