Design of View-Dependent Dynamic Text-Label Layouting in Interactive 3D Information-Rich Virtual Environments: A Compute-Budgeted Benchmark of Optimization Methods

Perceptually robust extended reality (XR) scenes should keep overlaid information legible and low-clutter under continuous viewpoint change within strict real-time budgets. Interactive 3D informationrich virtual environments (IRVEs) commonly use 3D labels (call-outs) that link short text to annotate objects. As users move and change viewpoint, callouts can become perceptually problematic due to occlusion, overlap, clutter, and ambiguous associations, while interactive deployments must compute layout adjustments under tight runtime budgets. In this work, we benchmark four dynamic meta-heuristic callout-arrangement methods: simulated annealing (SA), a non-degrading SA variant (H-SA), grey wolf optimization (GWO), and an extended GWO with neighborhood-style candidate injection (GWO-ext), together with a random baseline. Across callout densities and iteration budgets, we compare objective score, runtime, and an acceptable-solution rate. SA is most reliable and the only method reaching acceptable layouts at higher density (20 callouts) within tested budgets, while GWO-ext improves acceptability at lower density with more iterations.