A Hybrid Intelligent Decision-Making System for Navigation with Optimized Performance

A moving vehicle comes across some permanent components like roads and some variable components like other vehicles etc. Keeping these components at forefront, an automated vehicle needs to take astute and speedy decisions about its next move. In this research paper, we have proposed a hybrid model based on two soft computing techniques that is Neural Networks and Fuzzy Systems to optimize the performance of Navigation Systems in terms of speed and accuracy. The proposed model is an extension of authors previous work [ 1] in which the Best Feature Selection model was created to extract the most paramount features of navigation images. In that, it was observed that extracted feature set had direct impact on speed and accuracy of the working model as no resources were required to be spent on impertinent features. Our hybrid model performs three steps afore taking final decision that is either to move or stop the automated vehicle. The first step involves passing the outputs of Best Feature Selection Model and PCA as inputs through Convolutional Neural Network (CNN) deep learning model and obtaining the response in form of move or stop. Principal Component Analysis (PCA) technique was performed on the extracted feature set to improve our knowledge about input feature set by procreating new features. In our previous work [2], it was justified that Principal Component Analysis (PCA) when used in conjunction with Convolutional Neural Networks (CNN) yields better results as compared to standalone performance of PCA and CNN. It was withal analysed that PCA-CNN is going to give nearly same precision either we give ten percent training or we give high order training. The second step involves using one other variant of Neural Networks that is Re-current Neural Networks (RNN) for ascertaining out the bestest feature from all eleven features of Best Feature Selection Model. The factor with least Mean Square Error is used to calculate the decision in form of move or stop. The last step involves passing the outputs of CNN model and RNN model through Fuzzy Module. Here the Fuzzy rules are generated to take final decision for moving vehicle. The accuracy of this hybrid model came out be 89.28 percent which is far better than individual accuracies that is 72 percent and 60 percent of CNN model and RNN model respectively