We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.RO

Change to browse by:

cs

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Robotics

Title: POMP++: Pomcp-based Active Visual Search in unknown indoor environments

Abstract: In this paper we focus on the problem of learning online an optimal policy for Active Visual Search (AVS) of objects in unknown indoor environments. We propose POMP++, a planning strategy that introduces a novel formulation on top of the classic Partially Observable Monte Carlo Planning (POMCP) framework, to allow training-free online policy learning in unknown environments. We present a new belief reinvigoration strategy which allows to use POMCP with a dynamically growing state space to address the online generation of the floor map. We evaluate our method on two public benchmark datasets, AVD that is acquired by real robotic platforms and Habitat ObjectNav that is rendered from real 3D scene scans, achieving the best success rate with an improvement of >10% over the state-of-the-art methods.
Comments: Accepted at 2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)
Subjects: Robotics (cs.RO)
Cite as: arXiv:2107.00914 [cs.RO]
  (or arXiv:2107.00914v2 [cs.RO] for this version)

Submission history

From: Francesco Giuliari [view email]
[v1] Fri, 2 Jul 2021 09:05:40 GMT (11844kb,D)
[v2] Fri, 5 Nov 2021 08:38:10 GMT (11863kb,D)

Link back to: arXiv, form interface, contact.