Dr. Lu Qun from the Advanced Research Institute/School of Intelligent Manufacturing of Taizhou University, in collaboration with Professor Song Haiyu from Zhejiang University of Finance and Economics, jointly guided undergraduate students Lu Zedan, Xiang Houdong, and Yang Chengru to publish a research article in the Chinese Academy of Sciences Zone 1 Top journal IEEE Transactions on Systems, Man, and Cybernetics: Systems (IEEE TSMC). The paper is titled “Neuro-adaptive Safe Consensus Tracking Control for Pure-Feedback Nonaffine Multiagent Systems”. Among the authors, Dr. Lu Qun is the first author, Professor Song Haiyu is the corresponding author, and TU is the first author affiliation. TSMC is the flagship journal of the IEEE Systems, Man, and Cybernetics Society. It focuses on interdisciplinary fields such as systems engineering, artificial intelligence, and cybernetics. It is dedicated to publishing groundbreaking theoretical, methodological, and applied research achievements in these areas. Its current impact factor is 8.7.

Multi-agent systems have broad application prospects in fields such as UAV formation, intelligent transportation, and industrial automation. One of the core challenges is designing distributed control strategies that enable follower agents to safely and accurately track the leader’s trajectory under conditions of only local communication, unknown dynamic models, and external disturbances. The agent dynamics focused on in this research are of an unknown pure-feedback non-affine form, which renders traditional control methods based on precise models ineffective. Moreover, due to the non-linear manifestation of input variables, control design is particularly complex. To address this control challenge, this research effectively integrates multiple advanced control theories and tools:

● By utilizing the mean value theorem, the non-affine dynamics are transformed into a form suitable for the backstepping design framework, laying the foundation for systematic controller design.

● The introduction of dynamic surface control technology effectively avoids the inherent “computational explosion” problem in traditional backstepping design, significantly reducing the online computational burden of the controller;

● The use of radial basis function neural networks for online learning and approximation of unknown nonlinear functions and disturbances in the system endows the controller with strong adaptive and learning capabilities;

● By combining graph theory to handle distributed communication constraints and innovatively employing barrier Lyapunov functions, safety constraints are embedded in the controller design, ensuring that the tracking errors of all followers remain within preset safety boundaries.

This research not only provides a systematic design framework and solid theoretical guarantees for solving the consensus tracking problem of multi-agent systems with unknown non-affine dynamics, but also opens up new avenues for addressing safety control issues in a broader class of nonlinear systems through its technical approach of integrating neural network adaptation, dynamic surface control, and barrier Lyapunov functions.

Paper link: https://ieeexplore.ieee.org/document/11333876