Neural Networks Trained with Levenberg-Marquardt-Iterated Extended Kalman Filter for Mobile Robot Trajectory Tracking

Aissa, Ben Cherif; Chouireb, Fatima

doi:10.25103/jestr.104.23

Cited by 9 publications

(5 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The dynamic network structure revealed by time-varying brain neural network analysis may also provide new inspiration to develop biological neural networks, which may help establish more effective neural models to better simulate and understand human behaviors. In addition to these applications, one can also note that the kernel of ADTF is to solve the MVAAR model with a Kalman filter problem, which has been widely utilized in various research fields such as trajectory tracking [62] and neural network training [63]. In essence, our proposed L1-ADTF solved the Kalman filter problem in the L1-norm space, which may further resist the outlier influence during these applications and offer reliable results.…”

Section: Discussionmentioning

confidence: 99%

L1-norm based time-varying brain neural network and its application to dynamic analysis for motor imagery

Bore

et al. 2022

J. Neural Eng.

View full text Add to dashboard Cite

Objective. EEG-based motor imagery (MI) brain-computer interface offers a promising way to improve the efficiency of motor rehabilitation and motor skill learning. In recent years, the power of dynamic network analysis for MI classification has been proved. In fact, its usability mainly depends on the accurate estimation of brain connection. However, traditional dynamic network estimation strategies such as adaptive directed transfer function (ADTF) are designed in the L2-norm. Usually, they estimate a series of pseudo connections caused by outliers, which results in biased features and further limits its online application. Thus, how to accurately infer dynamic causal relationship under outlier influence is urgent. Approach. In this work, we proposed a novel ADTF, which solves the dynamic system in the L1-norm space (L1-ADTF), so as to restrict the outlier influence. To enhance its convergence, we designed an iteration strategy with the alternating direction method of multipliers (ADMM), which could be used for the solution of the dynamic state-space model restricted in the L1-norm space. Furthermore, we compared L1-ADTF to traditional ADTF and its dual extension across both simulation and real EEG experiments. Main results. A quantitative comparison between L1-ADTF and other ADTFs in simulation studies demonstrates that fewer bias errors and more desirable dynamic state transformation patterns can be captured by the L1-ADTF. Application to real MI EEG datasets seriously noised by ocular artifacts also reveals the efficiency of the proposed L1-ADTF approach to extract the time-varying brain neural network patterns, even when more complex noises are involved. Significance. The L1-ADTF may not only be capable of tracking time-varying brain network state drifts robustly but may also be useful in solving a wide range of dynamic systems such as trajectory tracking problems and dynamic neural networks.

show abstract

Section: Discussionmentioning

confidence: 99%

L1-norm based time-varying brain neural network and its application to dynamic analysis for motor imagery

Bore

et al. 2022

J. Neural Eng.

View full text Add to dashboard Cite

show abstract

“…Neural network architecture can be formed using two or more combined neurons to develop a multi-layer network [23]. Fig (3) represents an example of a multilayer architecture for a neural network [24]. The architecture of a neural networks consist of three layers, i.e., input layer which accepts system states or sensors real-time measurements, a hidden layer which is the intermediate layer of the network and output layer which is the model outputs or motors inputs.…”

Section: Neural Network Optimization Algorithmsmentioning

confidence: 99%

“…By using (21), (24) and (28), the weights updating rule of the Gauss-Newton algorithm can be illustrated as in below:…”

Section: Gauss-newton Algorithmmentioning

confidence: 99%

Supervised Neural Network Control of Real-Time Two Wheel Inverted Pendulum

Nabil

2020

Journal of Advanced Engineering Trends

View full text Add to dashboard Cite

This research paper investigates an intelligent control technique stabilizing a real-time model ofthe two-wheel inverted pendulum. The TWIP model is a highly non-linear, open-loop, and unstable system which makes control a challenge. Initially,a state-feedback controller that uses the dynamical system states and control signals to construct the precise control decision is used to stabilize the system. Later, Supervised Feed-Forward Neural Networks (SFFNN) based on back propagation Levenberg-Marquardt optimization algorithm are trained by using real-time measurements of system states and motors control signals from state-feedback controller stabilization. SFFNN control the two-wheel inverted pendulum better than a state-feedback controller.

show abstract

“…Although LM requires more memory than other algorithms, it is found to be efficient in minimizing sum of square errors between data points and functions. LM utilizes gradient descent or Gauss-Newton methods when parameters are far or close to their optimum values, respectively [10,11,12]. The LM algorithm is further expressed in the following equation:…”

Section: Fig 10 Artificial Neural Network (Ann) Modelmentioning

confidence: 99%

Characterization of a Mobile Waste-Robot: A Heuristic Method to Path Planning using Artificial Neural Networks

Corpuz

2019

IJRTE

View full text Add to dashboard Cite

In the field of mobile robotics, path planning is one of the most widely-sought areas of interest due to its nature of complexity, where such issue is also practically evident in the case of mobile robots used for waste disposal purposes. To overcome issues on path planning, researchers have studied various classical and heuristic methods, however, the extent of optimization applicability and accuracy still remain an opportunity for further improvements. This paper presents the exploration of Artificial Neural Networks (ANN) in characterizing the path planning capability of a mobile waste-robot in order to improve navigational accuracy and path tracking time. The author utilized proximity and sound sensors as input vectors, dual H-bridge Direct Current (DC) motors as target vectors, and trained the ANN model using Levenberg-Marquardt (LM) and Scaled Conjugate (SCG) algorithms. Results revealed that LM was significantly more accurate than SCG algorithm in local path planning with Mean Square Error (MSE) values of 1.75966, 2.67946, and 2.04963, and Regression (R) values of 0.995671, 0.991247, and 0.983187 in training, testing, and validation environments, respectively. Furthermore, based on simulation results, LM was also found to be more accurate and faster than SCG with Pearson R correlation coefficients of rx=.975, nx=6, px=0.001 and ry=.987, ny=6, py=0.000 and path tracking time of 8.47s.

show abstract

Neural Networks Trained with Levenberg-Marquardt-Iterated Extended Kalman Filter for Mobile Robot Trajectory Tracking

Cited by 9 publications

References 21 publications

L1-norm based time-varying brain neural network and its application to dynamic analysis for motor imagery

L1-norm based time-varying brain neural network and its application to dynamic analysis for motor imagery

Supervised Neural Network Control of Real-Time Two Wheel Inverted Pendulum

Characterization of a Mobile Waste-Robot: A Heuristic Method to Path Planning using Artificial Neural Networks

Contact Info

Product

Resources

About