Resumen

Safe and repeatable takeoff performance is critical in general aviation, especially at high altitude airports where density altitude and terrain constraints reduce safety margins. At Aeroclub of Colombia, instructors currently rely on qualitative observation and debriefing to judge whether students tracked the runway centerline efficiently during takeoff. This thesis develops and validates a Pilot Takeoff Efficiency Dashboard that transforms raw Garmin flight logs into quantitative metrics and visual feedback for instructors and students at a high-altitude runway. A Python-based data pipeline was implemented to import Garmin .csv files, clean and segment each flight, and reconstruct the aircraft trajectory in a runway-aligned reference frame. A Plotly/Dash dashboard overlays the flight path on a digital elevation model and provides synchronized plan, profile, and time-series views. The thesis introduces two novel geometric metrics for takeoff precision: • Mean Absolute Offset (MAO) – the time-average of the absolute lateral distance between the aircraft ground track and an ideal runway centerline over the takeoff roll. MAO, expressed in meters, captures how far the pilot typically flies from the centerline, regardless of direction. • Lateral Weave per Meter (LWM) – the cumulative lateral distance traveled (sum of absolute side-to-side motion) divided by the along-run distance. LWM, in m/m, quantifies how much the pilot “weaves” while accelerating, even if the final offset remains small. These metrics feed into two scalar performance scores: a restricted Pilot Takeoff Efficiency Score (PTES_restricted) and an adjusted score (PTES_adjusted). Both scores are normalized on a 0–140 scale where values below ~ 40 indicate weak alignment, around 80 –100 represent solid, proficient performance, and above 100 correspond to exceptionally precise and stable centerline tracking. PTES_restricted assigns a score of zero to any takeoff with at least one ±1 meter centerline crossing, making it a strict “no-crossings” benchmark. PTES_adjusted removes this hard veto and instead rewards low MAO and low weave while applying softer penalties to crossings, making it suitable for ranking and trend analysis. The system was applied to a dataset of 50 real training takeoffs collected in October–November 2025 on two aircraft. Typical performance was precise: MAO averaged 0.84 meters with a median of 0.70 meters, although a small number of flights exhibited offsets above 2 meters or up to eight crossings. PTES_restricted proved overly harsh: most flights collapsed to zero, limiting its usefulness beyond a binary standard. In contrast, PTES_adjusted produced a wide spread of scores (mean 55.5, range 0–136) and showed a strong, statistically significant relationship with MAO (R² ≈ 0.51, p ≪ 0.001), with additional contributions from lateral weave and crossings. The results show that simple geometric metrics derived from routine Garmin logs can reliably quantify takeoff precision at a high-altitude airport and support richer debriefs. The proposed dashboard enables instructors to identify best-practice flights, diagnose outliers, and monitor cohort-level performance over time. Future work should expand the dataset, incorporate wind and runway-condition data, and integrate individual pilot identifiers to study learning curves and instructor effects.