A homogeneous formalism for reasoning and learning in mobile robotics

Abstract

Autonomous systems must be able to perform both abstract reasoning and data interpretation. In order to bridge the formal gap between these two levels, a well-founded calculus based on a many-valued logic is introduced. This formalism preserves the ability to perform deductions on the exact abstract knowledge. In addition, it allows to reason on the uncertain knowledge, provided an appropriate data interpretation function is built through a learning process. An example of application to sonar-sensorized mobile robotics is presented.