Find minimum racks for 2D bin packing

You are given N machines. Machine i requires two types of resources: (x_i, y_i). Each rack has capacity (A, B) for the two resources, and a machine must be placed entirely within a single rack. Multiple machines can share a rack as long as the sum of their resource requirements in each dimension does not exceed the rack capacity. Return the minimum number of racks needed to place all machines. Notes: - All racks are identical. - You may assign machines to racks in any grouping/order. - The naive approach is brute force / bitmask DP over subsets. Can you design a more efficient algorithm (or improved DP) and analyze its time complexity under reasonable constraints (e.g., N up to ~20–25 for exact solutions, or larger if using approximation/heuristics)?