Adaptive Neuro Fuzzy Inference System ANFIS Based Multi Sensor Data Fusion for Improved Range Estimation in Robotic Navigation

Abstract

The importance of multi sensor data fusion has been increased tremendously over the past decade. The data fusion systems could provide a meaningful information, based on the data/information from different sources/sensors. The fusion process improves the performance of overall system in terms of better detection capabilities, reliability, fault tolerant nature, decision making capability etc. The data fusion systems were widely used for a variety of applications including medical diagnostics, surveillance, remote sensing, robotics etc. Probability theory, Evidence Theory, Soft computing methods such as Fuzzy Logic, Neural Networks etc. were explored well to address the research problems in the design of data fusion systems. This thesis presents the design and validation of soft computing based methods for the fusion of sensor data for range estimation in robotics. The thesis also reviews the state of art fusion approaches and its application in the design of intelligent systems. Sensor data fusion models were designed and evaluated, based on the range the data from Ultrasonic, Infrared, LIDAR sensors. The soft computing algorithms such as Neural Networks (NN), Fuzzy Inference Systems (FIS), Adaptive Neuro-Fuzzy Inference Systems(ANFIS), Support Vector Regression (SVR) methods etc. were used for the analysis. It was found that, the fusion exercise on sensor data could improve the overall performance of the system. The possibilities of improving the accuracy of a single sensor using the soft computing techniques such as NN, ANFIS, SVR, Long Short-Term Memory (LSTM) systems also explored in connection to the same. It will be possible to improve the accuracy of any sensor using the proposed technique. newline