Neural networks : a classroom approach

Neural networks with a good modeling capability have been used increasingly to predict and forecast solar radiation. Even diverse application of neural network has been reported in literatures such as robotics, pattern recognition, forecasting, power systems, optimization and social/psychological sciences etc. The models have categorized the review under three major performance schemes such as delay, number of neurons and activation function for establishment of neural network architecture. In each of these categories, we summarize the major applications of eight well recognized and often used neural network models of which the last two are custom based. The anticipated model are initiated and validated with 10 metrological parameters further in sub-categories. Evaluation of its accuracy associated with special flexibility of the model is demonstrated through the results based on parameter range. In summary, we conclude the best result showing that the delays, neuron, transfer function, model, parameters and RMSE errors are in range of 15 or 30, 10 or 20, tansig, Elman Back Propagation network, bulb point temperature or direct normal radiation, 9–10 and 25–35% training to the test cases. The review discloses the incredible view of using the neural networks in solar forecast. The work of other researchers in the ﬁeld of renewable energy and other energy systems is also reported which can be used in the future in the works of this field.

Solar radiation forecasting with multiple parameters neural networks

Improving positioning accuracy of vehicular navigation system during GPS outages utilizing ensemble learning algorithm

Position and orientation system (POS) is a key technology widely used in remote sensing applications, which integrates inertial navigation system (INS) and GPS using a Kalman filter (KF) to provide high-accuracy position, velocity, and attitude information for remote sensing motion compensation. However, when GPS signal is blocked, the POS accuracy will decrease owing to the unbounded INS error accumulation. To improve the reliability and accuracy of POS, this paper proposes a hybrid prediction method for bridging GPS outages. This method uses radial basis function (RBF) neural network coupled with time series analysis to forecast the measurement update of KF, resulting in reliable performance during GPS outages. In verifying the proposed hybrid prediction method, a flight experiment was conducted in 2011, based on a high-precision Laser POS. Experimental results show that the proposed hybrid prediction method is more effective than two other methods (KF and RBF neural network).

A Hybrid Prediction Method for Bridging GPS Outages in High-Precision POS Application

Aiming to improve the navigation accuracy during global navigation satellite system (GNSS) outages, an algorithm based on long short-term memory (LSTM) is proposed for aiding inertial navigation system (INS). The LSTM algorithm is investigated to generate the pseudo GNSS position increment substituting the GNSS signal. Almost all existing INS aiding algorithms, like the multilayer perceptron neural network (MLP), are based on modeling INS errors and INS outputs ignoring the dependence of the past vehicle dynamic information resulting in poor navigation accuracy. Whereas LSTM is a kind of dynamic neural network constructing a relationship among the present and past information. Therefore, the LSTM algorithm is adopted to attain a more stable and reliable navigation solution during a period of GNSS outages. A set of actual vehicle data was used to verify the navigation accuracy of the proposed algorithm. During 180 s GNSS outages, the test results represent that the LSTM algorithm can enhance the navigation accuracy 95% compared with pure INS algorithm, and 50% of the MLP algorithm.

A LSTM Algorithm Estimating Pseudo Measurements for Aiding INS during GNSS Signal Outages

Navigation systems used in recent days rely mainly on Kalman filter to fuse data from global positioning system (GPS) and the inertial navigation system (INS). In common, INS/GPS data fusion provides reliable navigation solution by overcoming drawbacks such as signal blockage for GPS and increase in position errors with time for INS. Kalman filtering INS/GPS integration techniques used in present days have some inadequacies related to the stochastic error models of inertial sensors, immunity to noise, and observability. This paper aims to introduce a new system integration approach for fusing data from INS and GPS utilizing artificial neural networks (ANN). A multi-layer perceptron ANN has been recently suggested to fuse data from INS and differential GPS (DGPS). Though the integrated system using multi-layer perceptron scheme improves the positioning accuracy, it has shortcomings like complexity with respect to the architecture of multi-layer perceptron networks and limitation of online training algorithm to provide real-time capabilities. This paper, therefore, proposes the use of an alternative ANN architecture. The proposed architecture is based on radial basis function (RBF) neural networks, which generally have simpler architecture and faster training procedures than multi-layer perceptron networks. The RBF-ANN module is trained to predict the INS position error and provide accurate positioning of the moving vehicle.

Integration of INS and GPS using radial basis function neural networks for vehicular navigation

An aircraft system mainly relies on a Global Positioning System GPS to provide accurate position values consistently. However, GPS receivers may encounter frequent GPS absence because of ephemeric error, satellite clock error, multipath error, and signal jamming. To overcome these drawbacks, generally a GPS is integrated with an Inertial Navigation System INS mounted inside the vehicle to provide a reliable navigation solution. INS and GPS are commonly integrated using a Kalman filter KF to provide a robust navigation solution. In the KF approach, the error models of both INS and GPS are required; this leads to the complexity of the system. This research work presents new position update architecture NPUA which consists of various artificial intelligence neural networks AINN that integrate both GPS and INS to overcome the drawbacks of the Kalman filter. The various AINNs that include both static and dynamic networks described for the system are radial basis function neural network RBFNN, backpropagation neural network BPN, forward-only counter propagation neural network FCPN, full counter propagation neural network Full CPN, adaptive resonance theory-counter propagation neural network ART-CPN, constructive neural network CNN, higher-order neural networks HONN, and input-delayed neural networks IDNN to predict the INS position error during GPS absence, resulting in different performances. The performances of the different AINNs are analyzed in terms of root mean square error RMSE, performance index PI, number of epochs, and execution time ET.

Performance analysis of various artificial intelligent neural networks for gps/ins integration

A novel approach to the integration of GPS and INS using recurrent neural networks with evolutionary optimization techniques

This paper aims to introduce a novel approach named IMM-UKF-TFS (Interacting Multiple Model-Unscented Kalman Filter-Two Filter Smoother) to attain positional accuracy in the intelligent navigation of a manoeuvring vehicle. Here, the navigation filter is designed with an Unscented Kalman Filter (UKF), together with an Interacting Multiple Model algorithm (IMM), which estimates the state variables and handles the noise uncertainty of the manoeuvring vehicle. A model-based estimator named Two Filter Smoothing (TFS) is implemented along with the UKF-based IMM to improve positional accuracy. The performance of the proposed IMM-UKF-TFS method is verified by modelling the vehicle motion into Constant Velocity-Coordinated Turn (CV-CT), Constant Velocity – Constant Acceleration (CV-CA) and Constant Acceleration-Coordinated Turn (CA-CT) models. The simulation results proved that the proposed IMM-UKF-TFS gives better positional accuracy than the existing conventional estimators such as UKF and IMM-UKF.

https://www.cambridge.org/core/services/aop-cambridge-core/content/view/6F8F1F7C3374A6AA9A357DC2664A604B/S0373463313000404a.pdf/div-class-title-imm-ukf-tfs-model-based-approach-for-intelligent-navigation-div.pdf

IMM-UKF-TFS Model-based Approach for Intelligent Navigation

Aircraft navigation relies mainly on Global Positioning System (GPS) to provide accurate position values consistently. However, GPS receivers may encounter frequent GPS outages within urban areas, where satellite signals are blocked. To overcome this drawback, generally GPS is integrated with inertial sensors mounted inside the vehicle to provide a reliable navigation solution. Inertial Navigation System (INS) and GPS are commonly integrated using a Kalman filter (KF) to provide a robust navigation solution, overcoming situations of GPS satellite signals blockage. This work presents New Position Update Architecture (NPUA) for GPS and INS data integration. NPUA uses Constructive Neural Network (CNN) for training and prediction. New dynamic learning algorithm (NDLA) has been developed for CNN to predict the INS position error and determine the accurate position of the moving aircraft during signal blockages in GPS. GPS and INS data are integrated using CNN in NPUA and the results are simulated. The output obtained using CNN is compared with the performance of Multilayer Feed Forward Network (MFNN). The CNN is found to have optimal topology when compared to MFNN. CNN has better learning ability, network constructing ability and accuracy when compared to MFNN.

M. Malleswaran

Papers

Integration of INS and GPS using radial basis function neural networks for vehicular navigation

Performance analysis of various artificial intelligent neural networks for gps/ins integration

A novel approach to the integration of GPS and INS using recurrent neural networks with evolutionary optimization techniques

IMM-UKF-TFS Model-based Approach for Intelligent Navigation

CNN based GPS/INS data integration using new dynamic learning algorithm