TY - JOUR
T1 - Fully Distributed Adaptive NN-Based Consensus Protocol for Nonlinear MASs
T2 - An Attack-Free Approach
AU - Lv, Yuezu
AU - Zhou, Jialing
AU - Wen, Guanghui
AU - Yu, Xinghuo
AU - Huang, Tingwen
N1 - Publisher Copyright:
© 2012 IEEE.
PY - 2022/4/1
Y1 - 2022/4/1
N2 - This article works on the consensus problem of nonlinear multiagent systems (MASs) under directed graphs. Based on the local output information of neighboring agents, fully distributed adaptive attack-free protocols are designed, where speaking of attack-free protocol, we mean that the observer information transmission via communication channel is forbidden during the whole course. First, the fixed-time observer is introduced to estimate both the local state and the consensus error based on the local output and the relative output measurement among neighboring agents. Then, an observer-based protocol is generated by the consensus error estimation, where the adaptive gains are designed to estimate the unknown neural network constant weight matrix and the upper bound of the residual error vector. Furthermore, the fully distributed adaptive attack-free consensus protocol is proposed by introducing an extra adaptive gain to estimate the communication connectivity information. The proposed protocols are in essence attack-free since no observer information exchange among agents is undertaken during the whole process. Moreover, such a design structure takes the advantage of releasing communication burden.
AB - This article works on the consensus problem of nonlinear multiagent systems (MASs) under directed graphs. Based on the local output information of neighboring agents, fully distributed adaptive attack-free protocols are designed, where speaking of attack-free protocol, we mean that the observer information transmission via communication channel is forbidden during the whole course. First, the fixed-time observer is introduced to estimate both the local state and the consensus error based on the local output and the relative output measurement among neighboring agents. Then, an observer-based protocol is generated by the consensus error estimation, where the adaptive gains are designed to estimate the unknown neural network constant weight matrix and the upper bound of the residual error vector. Furthermore, the fully distributed adaptive attack-free consensus protocol is proposed by introducing an extra adaptive gain to estimate the communication connectivity information. The proposed protocols are in essence attack-free since no observer information exchange among agents is undertaken during the whole process. Moreover, such a design structure takes the advantage of releasing communication burden.
KW - Adaptive attack-free protocol
KW - Consensus
KW - Distributed fixed-time observer
KW - Neural network (NN) approximation
UR - http://www.scopus.com/inward/record.url?scp=85098787882&partnerID=8YFLogxK
U2 - 10.1109/TNNLS.2020.3042821
DO - 10.1109/TNNLS.2020.3042821
M3 - Article
C2 - 33351766
AN - SCOPUS:85098787882
SN - 2162-237X
VL - 33
SP - 1561
EP - 1570
JO - IEEE Transactions on Neural Networks and Learning Systems
JF - IEEE Transactions on Neural Networks and Learning Systems
IS - 4
ER -