This open access book focuses on robot introspection which has a direct impact on physical
human-robot interaction and long-term autonomy and which can benefit from autonomous anomaly
monitoring and diagnosis as well as anomaly recovery strategies. In robotics the ability to
reason solve their own anomalies and proactively enrich owned knowledge is a direct way to
improve autonomous behaviors. To this end the authors start by considering the underlying
pattern of multimodal observation during robot manipulation which can effectively be modeled
as a parametric hidden Markov model (HMM). They then adopt a nonparametric Bayesian approach in
defining a prior using the hierarchical Dirichlet process (HDP) on the standard HMM parameters
known as the Hierarchical Dirichlet Process Hidden Markov Model (HDP-HMM). The HDP-HMM can
examine an HMM with an unbounded number of possible states and allows flexibility in the
complexity of the learned model and the development of reliable and scalable variational
inference methods. This book is a valuable reference resource for researchers and designers in
the field of robot learning and multimodal perception as well as for senior undergraduate and
graduate university students.
