These tests load the same pure logic used by the app. Distances use a Chebyshev transform on 10 cm cells, measured to occupied cell boundaries. Rectangles use cell-centre rasterization; paths use four neighbours. Reported widths have 10 cm grid resolution.
Seat access uses the mover’s footprint touching the chair’s outer edge, with its centre offset from the rasterized chair edge by half its width and rounded outward to a grid centre. Sampling can conservatively understate a gap by up to 10 cm (the 120 cm entrance samples as 110 cm). A blocked seat’s widest route uses the largest mover that can reach that approach. Literal centre-cell contact with an obstacle chair cannot accommodate the required mover widths.
Running tests…