Hierarchical Dirichlet Process Hidden Markov Models for abnormality detection in robotic assembly

The Hierarchical Dirichlet Process Hidden Markov model (HDP-HMM) is a Bayesian non parametric extension of the classical Hidden Markov Model (HMM) that allows to infer posterior probability over the cardinality of the hidden space, thus avoiding the necessity of cross-validation arising in standard EM training. This paper presents the application of Hierarchical Dirichlet Process Hidden Markov Models (HDP-HMM) to error detection during a robotic assembly task. Force sensor data is recorded for successful and failed task executions and manually labeled. An HDP-HMM is then fit to a set of training trials for each task execution outcome. We show how posteriors on the learned models could be used to recognize on-line deviation from expected behavior, thus allowing the robotic system to promptly react to task execution errors.