End-of-Life Disassembly as a Benchmark for Active Perception in Robotic Manipulation

Jan Baumgärtner, Malte Hansjosten, Karam Shaheen

Keywords: disassembly, benchmark, robotic manipulation

TL;DR: End of life products are a nice disassembly for active perception because finding out whats wrong with an object requires active interaction.

Abstract:

Active perception is central to robotic manipulation
under uncertainty, yet the field lacks benchmarks that
demand this coupling in realistic settings. We argue that endof-
life (EOL) disassembly naturally exhibits the properties
required of such a benchmark: hidden defects, occlusion, and
instance-specific variation that force a robot to gather information
through physical interaction. We formalize this view as a
POMDP and present a 3D-printable replica of an angle grinder,
designed as a benchmark for EOL disassembly. We validate the
benchmark by demonstrating a system that actively reasons
about stuck or missing parts from interaction feedback and
completes disassembly tasks under these uncertainties.

Email Sharing: We authorize the sharing of all author emails with Program Chairs.

Data Release: We authorize the release of our submission and author names to the public in the event of acceptance.

Submission Number: 12