Task planning in the context of trajectory learning is related to generation of a plan for reproduction of a demonstrated task. This chapter presents the task planning approaches, including Gaussian mixture regression (GMR), spline regression, locally weighted regression, and Gaussian process regression. Task planning with spline regression employs a set of trajectories key points, related to the transitions between the hidden states in the statistical models of Hidden Markov model (HMM) and conditional random field (CRF). Locally weighted regression is used for trajectory generation with tasks encoded via dynamic motion primitives (DMPs). Local learning methods approximate the training data locally by using a set of functions, as opposed to the global regression methods which aim to calculate a single function to fit the training data at a global level. Locally weighted projection regression utilizes linear univariate local models by projecting the input data into a local plane.
Task Planning
Robot Learning by Visual Observation ; 73-127
2017-03-10
55 pages
Article/Chapter (Book)
Electronic Resource
English
Preference Planning: A Reactive Approach to Robot Task Planning
British Library Conference Proceedings | 1993
|EVA Task Timing and Timeline Planning
British Library Conference Proceedings | 2007
|EVA Task Timing and Timeline Planning
SAE Technical Papers | 2007
|