Linear and nonlinear method based design of data fusion algorithms for advanced driver assistant systems

Data fusion can be defined as the processes of combining different types of sensor data into a common data type. Multisensor data fusion refers to the synergistic combination of sensory data from multiple sensors to provide reliable and accurate information. The challenge is how such huge data which is independently diverse can be fused. Hence there is a necessity for sophisticated and efficient data fusion techniques. Various approaches to address this problem have been found in literature and can be categorized broadly as a statistical method, generic Bayesian filter, and its derived filters, AI, deep learning, etc. In this paper, two different approaches and techniques to address the problem of sensor data fusion are presented and their comparative analysis is shown. Data fusion algorithms using a generic linear method (Kalman filter) and a nonlinear method (extended Kalman filter) have been designed in this work. Further, the challenges in a fusion of a variety of data, measured by different sensors, are addressed considering multiple-sensor-based autonomous vehicles as the application area for the current work.