For ICRA 2020 Paper: Grounding Language to Landmarks in Arbitrary Outdoor Environments.
This dataset has 1540 natural language to LTL pairings. The data is stored in a three-column CSV: Column 1 – Natural language expression Column 2 – LTL grounding Column 3 – City
The dataset has been cleaned such that natural language instructions collected from ambiguously formatted map representations are not included. The instructions in the dataset have not been modified.
Any questions should be directed to the Brown University H2R lab: h2r.tellex.lab@gmail.com