
		<paper>
			<loc>https://jjcit.org/paper/308</loc>
			<title>MEDIAN SPLIT AGAINST THE SURFACE AREA HEURISTIC FOR BROAD-PHASE COLLISION DETECTION UNDER PER-FRAME REBUILD: A PAIRED AND DIRECTLY TIMED COMPARISON</title>
			<doi>10.5455/jjcit.71-1787543958</doi>
			<authors>Hamzah Asyrani Sulaiman,Che Ku Nuraini Che Ku Mohd,Mohamad Lutfi  Dolhalit,Abdullah Bade,Iznora Aini Zolkifly</authors>
			<keywords>Broad-phase collision detection,Bounding volume hierarchy,Surface area heuristic,Median split,Per-frame rebuild</keywords>
			<views>22</views>
			<downloads>8</downloads>
			<received_date>24-Aug.-2026</received_date>
			<revised_date>  5-Oct.-2026 and 14-Oct.-2026</revised_date>
			<accepted_date>  15-Oct.-2026</accepted_date>
			<abstract>For broad-phase hierarchies rebuilt every frame, the submitted three-axis surface area heuristic performs 17.8 to 
46.0 percent fewer box tests than the object median yet takes 9 to 365 percent more net time throughout, up to 
2,000 objects; no crossover beyond is excluded. So does a one-axis heuristic. With a median split at eight objects 
or fewer, it led by 1.1 to 2.4 percent in 12 configurations, 11 when repeated and 10 in a fourth session, too small 
to claim. A split at the midpoint of the node's box was faster than the object median in all 30 density-controlled 
configurations, by 21 to 39 percent as implemented. A hierarchy built in Morton order, measured at a referee's 
request, was faster than both at 100 objects or more, mainly in construction. No heuristic was faster than the 
midpoint split, nor the variant than the Morton build.</abstract>
		</paper>


