Effective human-aware motion planning requires understanding how humans adapt their spatial behavior when sharing a workspace with collaborative robots. This study investigates human avoidance behavior during human-robot spatial interaction as a dual-factor process, distinguishing between (i) a static influence related to robot embodiment and mounting configuration, and (ii) a dynamic influence related to trajectory characteristics. We conducted a repeated-measures experiment (N = 48) in which participants interacted with a robot executing trajectories with varying approach directions (e.g., head-on vs. parallel) and curvature profiles (Rectilinear vs. Curvilinear), as well as different mounting configurations (table-mounted vs. shoulder-mounted). Human responses were evaluated using continuous kinematic tracking and subjective workload assessment (NASA-TLX). The results indicate systematic differences in human spatial behavior depending on trajectory type and interaction geometry. Rectilinear trajectories were associated with increased subjective workload, altered movement timing, and expanded spatial clearance behavior, particularly in head-on interaction conditions. In contrast, Curvilinear trajectories tended to support more consistent human motion patterns during shared workspace negotiation, despite occupying larger spatial volume. Additionally, robot mounting configuration influenced perceived and behavioral safety margins, with table-mounted setups leading to increased proxemic distance even in the absence of explicit task disruption. This work provides an empirical characterization of how static and dynamic aspects of robot behavior jointly influence human spatial adaptation in shared environments, supporting the design of ergonomically informed human-robot interaction systems. This work is complemented by a parallel study focusing on continuous kinematic structure and biomechanical adaptation in the same experimental setting (see Related Works). Note: This manuscript is a preprint version of a paper currently under peer review at Advanced Robotics Research (Wiley).
Kaya et al. (Tue,) studied this question.