SPAR: Automated Static Placement of Virtual Content via Ergonomic‐Aware Multi‐Objective Optimization

Abstract Automated virtual content placement is a core challenge in augmented reality (AR) experience generation. The virtual content are typically categorized into object‐anchored content, which remains fixed in the physical environment, and user‐anchored content, which adapts dynamically to user context. While adaptive interfaces have been extensively studied, generating static, object‐anchored layouts, such as labels for machinery or museum exhibits, remains underexplored. Existing methods either depend on manual design, which is time‐intensive and requires expertise, or on rule‐based optimization, which often ignores ergonomic factors during user interaction. In this paper, we focus on the problem of static layout generation and propose SPAR , a novel two‐stage framework for S tatic P lacement of virtual content in AR applications. In the first stage, we estimate user viewpoints and shoulder joints by placing human models around the target object. In the second stage, we formulate and solve an ergonomic‐aware multi‐objective optimization problem to determine the final placement configuration. To achieve this, we propose a new metric, Extended Consumed Endurance (ECE), to evaluate arm fatigue during mid‐air interactions, and integrate ECE as a loss function into the optimization framework to enhance interaction‐centric layout generation. We conducted a comprehensive evaluation of both the placement framework and the ECE metric. Results demonstrate that SPAR outperforms baseline methods in generating arrangements aligned with human interaction patterns. ECE exhibits comparable evaluation capabilities to a widely adopted subjective fatigue metric (Borg CR10, (0.769)) and a recent state‐of‐the‐art computational metric (NICER, (0.812)). The SPAR source code is publicly available for academic research at https://github.com/Evenwang521/SPAR .