Real-Time Performance Optimization in a 3D PlayCanvas Engine-Based Web Configurator

Sammanfattning

Real-time 3D rendering in web browsers introduces severe CPU bottlenecks due to JavaScript’s single-threaded execution and WebGL API draw call overhead. This thesis investigates and aims to resolve these performance constraints within a commercial PlayCanvas-based 3D warehouse configurator intended to support ~100 000 pallet meshes. Through a quantitative, full-factorial experimental methodology, this research evaluates the impact of targeted architectural optimizations; specifically, hardware instancing, geometric mesh merging, and substituting continuous ammo.js physics with lightweight Axis-Aligned Bounding Boxes. The results indicate a reduction in overhead – suggesting a shift of the primary performance bottleneck from the CPU-bound software layer to the hardware GPU. Ultimately, these findings contribute to a basis for the feasibility of complex browser-based 3D rendering and the applicability of relevant foundational concepts in this context.