The diversity of the solutions employed was impressive
at a hardware, software and algorithms level. They ranged
from large, single robot arms to multiple small robots each
assigned to one bin on the shelf, from simple suction cups
to anthropomorphic robotic hands, and from fully reactive
approaches to fully deliberative sense-plan-act approaches. In
surveying the details of each team’s approach and questioning
them on what they learned from the experience, we hope to
extract trends that help us (1) understand how to eventually
solve the problem, and (2) discover what future robotics
research directions are most promising for solving the general
problems of perception, manipulation, and planning.