Automated Alignment of Specifications of Everyday Manipulation Tasks. Tenorth, Zeigltrum, Beetz. IROS 2013

  1. Discusses robot task transfer
  2. Idea is to use algorithms for sequence alignment from bioinformatics on action specifications.  Extensions include:
    1. “methods for the comparison of complex sequence elements”
    2. taking semantic similarity into account
    3. aligning descriptions at different levels of granularity
  3. Also, how do you turn natural language instructions into something a robot understands?
  4. Although DNA is long, it only has 4 components, action descriptions “have a complex structure that relates action classes to the involved objects, tools, and locations, and can also be hierarchically structured and composed of sub-actions.” So extensions are needed for this
  5. In earlier work, they took instructions from wikihow and parsed them and embedded them in their global robot database
  6. For bioinformatics, there are 2 major algorithms, based on DP
    1. Needleman-Wunsch: does global alignment
    2. Smith-Waterman: optimal local alignment
    3. <Not taking notes on these>
  7. Issue of multiple ways of annotating the same behavior, with different nouns and verbs, which makes alignment hard
  8. Frequent sub-sequences are identified and stored as a higher-order behavior
  9.  Some nice results

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: