Rule-based Contact Monitoring Using Examples Obtained by Task Demonstration

This paper presents a new rule-based hybridsystem approach to contact monitoring. The rules are formulated in terms of temporal sequences of the contact force and they recognize temporal patterns of force associated with contact states and transitions. The rule-base is built using inductive learning techniques on force data obtained by human demonstration. This approach is suitable for monitoring robotic as well as human tasks. An advantage of this approach is that it allows process monitors for different tasks to be built quickly and easily by learning new sets of rules from a demonstration of the tasks. Experimental results are presented to demonstrate the effectiveness of this approach. 1 Introduction Robotic systems are prone to errors due to uncertainties in the model of the process and of the environment. This makes process monitoring an important part of a robotic system. Process monitoring is based on continuously observing and interpreting the data from sensors that report on..