The size of datasets is growing exponentially as information technology advances, and it is becoming more and more crucial to provide efficient learning algorithms for neural networks to handle massive amounts of data. Due to their potential for handling huge datasets, feed-forward neural networks with random weights (FNNRWs) have drawn a lot of attention. In this paper, we introduced an efficient feed-forward neural network scheme (FNNS) for processing massive datasets with random weights. The FNNS divides large-scale data into subsets of the same size, and each subset derives the corresponding submodel. According to the activation function, the optimal range of input weights and biases is calculated. The input weight and biases are randomly generated in this range, and the iterative scheme is used to evaluate the output weight. The MNIST dataset was used as the basis for experiments. The experimental results demonstrate that the algorithm has a promising future in processing massive datasets.

/pdf/fnns-an-effective-feedforward-neural-network-scheme-with-tcn424et.pdf

FNNS: An Effective Feedforward Neural Network Scheme with Random Weights for Processing Large-Scale Datasets

Precision during the guided terminal flight in short-range rockets is a key factor that must be improved using several methodologies. Traditionally, inertial navigation systems allowed to unbind the accuracy concerning the range. Unfortunately, in short-range rockets, the response of the inertial-based controller has to be fast enough to correct the trajectory in a short time. This fact implies that to achieve high precision, either the use of systems too large to be implemented in portable rockets is required, or systems that are too expensive concerning the total cost of the rocket. For these systems, artificial intelligence-based guidance tactics could improve the precision, since once the network is trained, it is not necessary to know the dynamics of the vehicle, and the network itself can react quickly, simplifying the system and reducing economic costs using relatively simple electronics. The neural networks are trained using a nonlinear-dynamics model based on a simulated flight dynamics and verified with real flight data. The simulation results show that the proposed method works effectively in a six-degree-of-freedom simulation environment, with excellent accuracy and robustness to parameter uncertainty. The appropriateness of the closed-loop performance is validated using Monte Carlo analysis across a wide range of uncertainty scenarios.

Neural Network-Based Controller for Terminal Guidance Applied in Short-Range Rockets

Observer-based optimal control method combination with event-triggered strategy for hypersonic morphing vehicle

Precision during the guided terminal flight in short-range rockets is a key factor that must be improved using several methodologies. Traditionally, inertial navigation systems allowed to unbind the accuracy concerning the range. Unfortunately, in short-range rockets, the response of the inertial-based controller has to be fast enough to correct the trajectory in a short time. This fact implies that to achieve high precision, either the use of systems too large to be implemented in portable rockets is required, or systems that are too expensive concerning the total cost of the rocket. For these systems, artificial intelligence-based guidance tactics could improve the precision, since once the network is trained, it is not necessary to know the dynamics of the vehicle, and the network itself can react quickly, simplifying the system and reducing economic costs using relatively simple electronics. The neural networks are trained using a nonlinear-dynamics model based on a simulated flight dynamics and verified with real flight data. The simulation results show that the proposed method works effectively in a six-degree-of-freedom simulation environment, with excellent accuracy and robustness to parameter uncertainty. The appropriateness of the closed-loop performance is validated using Monte Carlo analysis across a wide range of uncertainty scenarios. 

Entry trajectory optimization for hypersonic vehicles based on convex programming and neural network

An online generation method of ascent trajectory based on feedforward neural networks

Asymmetric integral barrier Lyapunov function-based dynamic surface control of a state-constrained morphing waverider with anti-saturation compensator

This paper investigates the attitude control problem for underactuated body-flap hypersonic vehicles (BFHSVs) with mixed disturbances. First, the control-oriented model for BFHSV is introduced. Then, an improved finite-time super twisting disturbance observer (STDO) is designed. Finite-time convergence of estimate error and smoother inputs are achieved. Meanwhile, a parametric command method is introduced to calculate the differential of inputs which can enhance the dynamic response of the closed-loop system. Subsequently, the virtual control signal is derived by a second-order filter to avoid the differential explosion problem. The overall stability of the closed-loop system is demonstrated by applying the Lyapunov stability theory. Finally, the performance of the proposed control scheme is evaluated through extensive and comparative numerical simulations under multiple disturbances.

/pdf/finite-time-super-twisting-disturbance-observer-based-1840ldwi.pdf

Pei Dai

Papers

An online generation method of ascent trajectory based on feedforward neural networks

Asymmetric integral barrier Lyapunov function-based dynamic surface control of a state-constrained morphing waverider with anti-saturation compensator

Entry trajectory optimization for hypersonic vehicles based on convex programming and neural network

Finite-Time Super Twisting Disturbance Observer-Based Backstepping Control for Body-Flap Hypersonic Vehicle