A new POMDP solver called GNPF-kCT uses belief tree reuse, neural process filtering, and k-center clustering to achieve faster and more reliable 3D object search than baselines in simulations and real office tests.
The explore action allows the robot to adjust its configuration, change its field of view, and make an initial observation of the object
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.RO 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
POMDP-based Object Search with Growing State Space and Hybrid Action Domain
A new POMDP solver called GNPF-kCT uses belief tree reuse, neural process filtering, and k-center clustering to achieve faster and more reliable 3D object search than baselines in simulations and real office tests.