Objective This paper presents classifications of m ental tasks based on EEG signals using an adaptive Radial Basis Function (RBF) n etwork with optimal centers and widths for the Brain-Computer Interface (BCI) s che...Objective This paper presents classifications of m ental tasks based on EEG signals using an adaptive Radial Basis Function (RBF) n etwork with optimal centers and widths for the Brain-Computer Interface (BCI) s chemes. Methods Initial centers and widths of the network are s elected by a cluster estimation method based on the distribution of the training set. Using a conjugate gradient descent method, they are optimized during train ing phase according to a regularized error function considering the influence of their changes to output values. Results The optimizing process improves the performance of RBF network, and its best cognition rate of three t ask pairs over four subjects achieves 87.0%. Moreover, this network runs fast du e to the fewer hidden layer neurons. Conclusion The adaptive RB F network with optimal centers and widths has high recognition rate and runs fas t. It may be a promising classifier for on-line BCI scheme.展开更多
This paper proposes a control method based on adaptive radial basis function neural networks(RBFNN)and dynamic inversion control for unmanned receiver to address the nonlinearities and uncertainties in the autonomous ...This paper proposes a control method based on adaptive radial basis function neural networks(RBFNN)and dynamic inversion control for unmanned receiver to address the nonlinearities and uncertainties in the autonomous aerial refueling docking process.Following the principle of time-scale separation,the attitude control of the unmanned receiver is divided into fast and slow loops.A dynamic inversion method is employed to design the attitude loop controller,and an adaptive RBFNN system is designed to compensate for system model errors and external disturbances.The efficacy and stability of the designed control method were validated through simulation experiments conducted during autonomous aerial refueling docking.The results indicate that the autonomous aerial refueling docking system based on adaptive RBFNN dynamic inversion control exhibits robustness and adaptability,enabling reliable refueling operations in complex aerial environments and providing reliable support for long-duration unmanned missions.展开更多
基金ThisworkwassupportedbytheNationalNaturalScienceFoundationofChina (No .3 0 3 70 3 95 )
文摘Objective This paper presents classifications of m ental tasks based on EEG signals using an adaptive Radial Basis Function (RBF) n etwork with optimal centers and widths for the Brain-Computer Interface (BCI) s chemes. Methods Initial centers and widths of the network are s elected by a cluster estimation method based on the distribution of the training set. Using a conjugate gradient descent method, they are optimized during train ing phase according to a regularized error function considering the influence of their changes to output values. Results The optimizing process improves the performance of RBF network, and its best cognition rate of three t ask pairs over four subjects achieves 87.0%. Moreover, this network runs fast du e to the fewer hidden layer neurons. Conclusion The adaptive RB F network with optimal centers and widths has high recognition rate and runs fas t. It may be a promising classifier for on-line BCI scheme.
基金supported by the National Natural Science Foundation of China under Grant 62473039the Joint Fund of Ministry of Education for Equipment Pre-Research under Grant 8091B03032303Beijing Nova Program 20240484561
文摘This paper proposes a control method based on adaptive radial basis function neural networks(RBFNN)and dynamic inversion control for unmanned receiver to address the nonlinearities and uncertainties in the autonomous aerial refueling docking process.Following the principle of time-scale separation,the attitude control of the unmanned receiver is divided into fast and slow loops.A dynamic inversion method is employed to design the attitude loop controller,and an adaptive RBFNN system is designed to compensate for system model errors and external disturbances.The efficacy and stability of the designed control method were validated through simulation experiments conducted during autonomous aerial refueling docking.The results indicate that the autonomous aerial refueling docking system based on adaptive RBFNN dynamic inversion control exhibits robustness and adaptability,enabling reliable refueling operations in complex aerial environments and providing reliable support for long-duration unmanned missions.
