Liquid-Handling Robot vs Human Pipetting: Precision Statistics (2D)

Assay Setup

Target volume 200 µL

Session length (trials) 96

Dispense rate 8 /s

Precision Model

Robot base precision (CV) 0.40%

Human baseline precision (CV) 2.50%

Human fatigue intensity 1.0×

Live Readout — Robot

n / mean

0 · — µL

SD / CV%

— / —

Live Readout — Human

n / mean

0 · — µL

SD / CV%

— / —

Robot bias

— µL

Human bias

— µL

How it works

Each trial draws a dispensed volume from a Gaussian around the target, using the Box–Muller transform: V = target + bias + σ·z, z ~ N(0,1). The robot's σ and bias are fixed for the whole session (mechanical repeatability). The human's σ and bias grow with session progress f = trial / sessionLength — a simple linear fatigue model:

σ_human(f) = target·CV_base/100 · (1 + fatigue·1.5·f)
bias_human(f) = target·(0.003 + fatigue·0.02·f)
σ_robot = target·CV_robot/100   (constant)
bias_robot = −0.002·target      (constant)

CV% (coefficient of variation) = sample SD ÷ sample mean × 100 — the same metric a real lab QC report uses to judge pipetting reproducibility (USP/ISO 8655 guidance typically wants manual pipetting CV% under ~2-3% and automated liquid handlers well under 1%). Sample SD uses the n−1 (Bessel-corrected) denominator.

Histogram — running distribution of dispensed volumes, robot vs human, rebinned every trial.
Error-vs-trial — each trial's deviation from target plotted against trial index; the human series visibly widens as the session progresses.
CV% trend — a trailing-window CV% recomputed every few trials, showing the robot staying flat while the human's drifts upward with fatigue.

Robot Human Target volume