Multisensor fusion of environment measures using Bayesian Networks

Abstract

Autonomous mobile robots usually require a large number of sensor types and sensing modules. There are different sensors, some complementary and some redundant. Integrating the sensor measures implies several multisensor fusion techniques. These techniques can be classified in two groups: low level fusion, used for direct integration of sensory data; and high level fusion, which is used for indirect integration of sensory data. We have developed a system to integrate indirect measures of different sensors. This system allows us to use any type of sensor which provides measures of the robot's environment It Is designed as a Belief Bayesian Network. The method needs that the user creates a low level fusion module and an interface between that module and our fusion system.