Data Mining and Fusion

Autonomous systems use measurements from proprioceptive and exteroceptive sensors to detect and track obstacles through the detection and tracking of moving objects (DTMO) approach, as well as to localize and plan a trajectory through the simultaneous localization and mapping approach. DTMO allows the detection and tracking of obstacles. Data fusion represents the scientific methods that make it possible to make indicators more reliable by aggregating data from various sources. This chapter presents three types of data fusion: redundant fusion, complementary fusion and cooperative fusion. It is customary to classify the types of data fusion according to levels of abstraction: low level, medium level or high level. Bayesian theory is one of the first approaches used to achieve fusion with decision-making. Mastering the reliability of obstacle detection is fundamental for the certification of the system and its implementation on commercial vehicles.