• Paper:

NOTES:

  • David Paulius and George Konidaris are joint last authors (equal advising).
  • This paper introduces the object scouting problem: a way of tackling different types of uncertainty (spatial uncertainty and state uncertainty) related to mobile manipulation and long-horizon task execution.
  • The object scouting problem is modelled using the locally observable Markov decision process (LOMDP) previously introduced by Merlin et al. 2024 .
  • We introduce a new planner specifically designed for object scouting known as SPOP (scouting partial-order planner), which is based on the idea of least commitment planning by Weld 1994 .
  • We show how this planner is better suited for planning to find task-relevant objects and resolve the plans to incorporate these objects than the previous LOMDP planner.
    • These are specifically highlighted in the AI2THOR simulator environment.

Citation:

M. Merlin, Z. Yang, G. Konidaris*, and D. Paulius* (2025). “Least Commitment Planning for the Object Scouting Problem”. In: Proceedings of the 2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), IEEE.