Deriving Expert System Expressions from Robotic Assembly Feedback

G. Lowe and B. Shirinzadeh (Australia)


autonomous, robotics, expert system, automatic detection.


Detecting the progress of part assembly and transforming this progress into symbolic language will support robot systems in decision making. This paper outlines an ap proach for autonomous conversion of measured properties into suitable input data for Reinforcement Learning (RL) and Knowledge-Base Self-Learning (KBSL) systems. The detection of part assembly requires an under standing of what to measure, and how to measure. The `what to measure', is defined by part assembly in terms of physical and functional states. The physical state involves measurement of the assembly in terms of pre, concurrent, and post operations to assess achievement relative to phys ical goals. The functional state tests operation of the part once each part is assembled. The `how to measure' prob lem focuses on physical state measurement with cameras, fibre optic lasers, force sensors, and sound. There measurements are translated into language us ing an expert system simulator.

Important Links:

Go Back