ICGI is the long-running venue for learning formal grammars and automata from data. The 2012 edition included a sequence-prediction challenge that asked entrants to model structured symbol streams from limited samples.
Grammatical inference is unfashionable relative to deep learning, but it answers a question robotics still cares about: how to learn a compact, inspectable model of a process from few examples.
Task graphs, behaviour trees and process models in modern industrial robots are all structured-sequence problems in this lineage.
This page preserves a historic coral-lab.org address. The original files are held offline; if you need them for citation or attribution, contact Coral Labs.