Obstacle Detection in an Unstructured Industrial Robotic System: Comparison of Hidden Markov Model and Expert System

This paper presents a comparison of two approaches for detecting unknown obstacles inside the workspace of an industrial robot using a laser rangefinder for 2-D measurements. The two approaches are based on Expert System (ES) and Hidden Markov Model (HMM). The results presented in the paper demonstrate that both approaches are able to correctly detect and classify unknown objects. The ES is characterised by low computational requirements and an easy setup when relatively few known objects are to be included inside the workspace. HMMs are characterised by a higher flexibility and the ability to handle a larger amount of known objects inside the workspace. Another significant benefit of the HMM approach taken in this paper, in contrast to voice recognition, is the fact that the learnt parameters of the HMMs have physical meaningful geometrical interpretations.