Learning Sequences of Controllers for Complex Manipulation Tasks
Jaeyong Sung, Bart Selman, Ashutosh Saxena · 2013
Many tasks in human environments require performing a sequence of complex navigation and manipulation tasks. In unstructured human environments, the locations and conguration of the objects can change in unpredictable ways. This requires a high-level planning strategy that is robust and exible in an uncertain environment. We propose a novel dynamic planning strategy, which can be trained from a set of example activities. High level activities are expressed as a sequence of primitive actions or controllers (with appropriate parameters). Our planning model synthesizes a universal strategy, where the a suitable next action is selected based on the current state of the environment. By expressing the environment using sets of attributes, the approach generalizes well to unseen scenarios. By unfolding our planning strategy into a Markov Random Field approximation, we can eectively train parameters using a maximum margin learning strategy. We provide a detailed empirical validation of our overall framework demonstrating successful plan strategies for a variety of tasks.