Constraint Aware Handover Scheduling Under Intermittent Occlusion

Abstract

Intermittent visibility creates a scheduling dilemma in human–robot handover: an early approach can intersect an uncertain reach, while a universal waiting buffer can accumulate queueing delay. This study examines that dilemma with a fully specified synthetic event simulator rather than robot experiments. A finite candidate scheduler chooses an approach time and one of two abstract routes using an observation-age uncertainty margin and explicit lateness cost. Four transparent policies are compared on 100 simulated runs containing 48 synthetic requests each, with common latent perturbations across policies. In these simulated data, the occlusion-aware policy reduces precontact threshold violations from 21.54% for mean-only scheduling to 2.92%, while increasing mean request-to-completion time from 1.899 to 2.983 s. An unconditional detour and waiting buffer reaches 0.77% simulated violations but incurs 8.206 s mean flow time in the deliberately congested finite queue. Sensitivity analyses expose uncertainty miscalibration, occlusion-duration effects, and candidate-budget tradeoffs. A constructed hidden-bias trajectory violates the threshold despite satisfying the scheduler's forecast margin, establishing the absence of a physical safety guarantee. The contribution is an auditable systems argument about timing, information, and queueing under stated synthetic assumptions; it supplies no evidence of real-world performance or human acceptability.