A tractor pulls an implement of width W across the field in parallel passes spaced W apart. If the driver — human or auto-steer — held a perfectly straight line, every pass would exactly abut the next with zero gap and zero overlap. In reality the vehicle wanders sideways off its planned line by a cross-track error (CTE) that depends on the guidance system:
CTE random walk (per row j along a pass):
e[j] = 0.85·e[j-1] + N(0, σ_eff) (mean-reverting: driver/auto-steer keeps correcting back)
σ_eff = σ_base(guidance) · (1 + max(0, v−6)·0.08) v = travel speed, km/h
σ_base: manual ≈ 0.25 m · WAAS GPS ≈ 0.08 m · RTK GPS ≈ 0.02 m
Each field cell is rasterised and counted for how many pass-swaths actually cover it:
covered(cell, pass i) = |x_cell − (center_i + e_i[row])| ≤ W/2
count(cell) = Σ_i covered(cell, i)
Skip % = cells with count = 0 ÷ total cells
Overlap % = cells with count ≥ 2 ÷ total cells
Waste % = Σ max(count−1, 0) ÷ total cells (extra chemical/seed dosed on double-covered ground)
- Guidance sets σ_base — this is the single biggest lever: RTK auto-steer holds the line roughly 10× tighter than a human driver.
- Implement width — a wider boom needs fewer passes, but the same CTE eats a larger fraction of each swath, so overlap/skip % actually worsens for very wide gear on rough ground.
- Travel speed — faster driving leaves less time to correct drift, inflating σ_eff for every guidance mode.
- Field width — sets how many passes the field needs; more passes means more edges where overlap and skips accumulate.
Real-world relevance: this is exactly why RTK auto-steer is standard on modern precision-ag equipment — cutting overlap from ~8–12% (manual) to under 1% pays for the hardware in fertilizer and seed savings alone within a season or two.