Mosaic helped a commercial airline integrate custom machine learning to predict departure runway and taxi-out time, cut operational costs, and improve the customer experience, aiding in reducing fuel emissions with AI.

Take Our Content to Go


The ability of the aviation industry to reduce delays on the ground during the departure process can lead to significant savings in the operational budget and reductions in environmental impact for both commercial and cargo airlines alike. In 2022, U.S. passenger airlines’ average cost of aircraft taxi delays was $80.52 per minute.  Yet, predicting taxi-out time is a complex process due to the uncertainties associated with factors under and out of the airlines’ control. Any effective machine learning model must factor in airport surface traffic, downstream traffic restrictions, runway configurations, weather, and air traffic control heuristics.

As a leader in aviation analytics with deep experience delivering cutting-edge artificial intelligence and machine learning solutions to aviation customers, Mosaic Data Science was more than capable of developing a custom solution to this legacy industry problem. Mosaic worked with one of the world’s largest commercial aviation outfits to fill their data science talent gap and design and deploy machine learning models that would predict departure runway and taxi-out time, cutting operational costs while improving the customer experience and sustainability efforts. The goal was to prove the feasibility of reducing fuel emissions with AI.


Taxi-out time is referred to as the duration of time from when an aircraft “pushes back” from its gate to when it takes off from the runway. Mosaic worked with a leading airline that delivered taxi-out time predictions to its pilots via Aviobook, a flight bag application that integrates aviation capabilities and information currently spread across many different apps into one place.

The taxi-out time predictions would be able to help pilots decide when to “single-engine taxi,” which refers to taxiing most of the way to the runway with only a single-engine turned on, turning on the second engine just a few minutes before take-off. When taxi-out times are relatively long, single-engine taxiing can save fuel. The company anticipated that the fuel savings from additional single-engine taxiing attributable to the predictions provided in the app would be over $3 million per year—sufficient on their own to justify the entire expense of developing the app.

The single-engine taxi decision is typically made by pilots within an hour of pushback. However, the airline wanted to deliver predictions up to four hours before the expected pushback time, so flight dispatchers could better anticipate taxi-out times and plan consistently with corresponding likely pilot single-engine taxi decisions.

Taxi Time ML Development

Developing the real-time data and software infrastructure to support these models and establishing appropriate connectivity between Mosaic and the customer was a tremendous effort that, fortunately, could stand on the giant shoulders of ATD-2.

Mosaic needed to lay out a development plan that quickly delivered value to the airline and provided quick wins along the way. First, the data science team needed to develop a taxi-out time prediction model that factored in the variability of airport surface operations. Once this model was deemed acceptable, the team designed a departure runway prediction model and deployed an interface to integrate predictions into the app, getting insights into decision-makers’ hands.

Early in model development, Mosaic trained models using features derived from weather data (LAMP forecasts). While these proved helpful to the models, our team eventually learned that this weather data would not be available in the deployed system, so these features must be disregarded for now.

However, Mosaic also left out an essential element during the initial taxi-out time model development because we didn’t realize that it would be readily available: predicted runways. Once the predicted runway model was ready, its predictions could serve as essential features for the taxi-out time prediction model.

Throughout the model development process, Mosaic collaborated with the airline customer to ensure the accuracy of the predictions and the project’s success. There are several factors to model, so Mosaic implemented a feedback loop with the airline to discuss which variables were important in the predictive process, aligning on the outcomes that matter to pilots.

Working Through User Acceptance

Mosaic encountered two ML-creeps in scope related to user trust in model predictions.

The first was to provide a vaguely defined “percent confidence” with the predicted runway and taxi-out time for each flight. The airline had developed requirements for the new Aviobook app with the help of a working group of pilots; that group indicated that a pilot’s single-engine taxi decision should be based not only on the predicted taxi-out time but also on the trustworthiness of that particular prediction.

Mosaic developed confidence intervals on the predictions to address this concern, using similar techniques described here.

The second creep in scope is also related to the trustworthiness or interpretation of model predictions manifested in a request that model input features be provided alongside model predictions. In particular, the airline requested that the “future demand” and “future capacity” shown in the app screenshot above be provided by Mosaic and based on the same data processing that prepares data for the model. This effort ensures that these numbers are consistent with the predicted taxi-out times and help users identify cases when the model works with forecasted traffic demand data that may not align with other forecasts.


The runway prediction models gave the airline the insights it needed to make more efficient and sustainable decisions and communicate them to its pilots. In addition, the airline was more capable of reducing ground delays and providing a better customer experience. Finally, the airline’s planes operated more sustainably by avoiding burning off as much jet fuel, reducing carbon emissions with single-engine taxi support. The departure process was streamlined so that the airline listens to what the data and the real-time situation told them.

These ML models offer similar benefits to an ATD-2 project Mosaic worked on, and NASA has calculated the following environmental outputs of implementing this technology:

Looking ahead, Mosaic will continue to deploy and update these machine learning models to serve the commercial airline best. The solutions developed will aid greatly in reducing fuel emissions with AI.

Mosaic has developed models and supporting software to predict departure runway and taxi time for all mainline departures at three important airports (EWR, ORD, & SFO). Mosaic will roll these out at 27 additional airports in the coming months.