Airport workforce optimization
October-November 2019
About the company
This Challenge was proposed by one of the major Latin American airline companies. It operates all around the world and covers more than 100 countries. This challenge dealt with workforce scheduling in El Dorado airport (Bogotá), one of the company's most important hubs.
The Challenge
This workforce management problem combines two optimization problems.
On one hand, there is a scheduling problem, where many decisions have to be determined: how to design the shifts (start, finish, break time) to fulfill the demand and all the constraints related to corporate policies, legislation and staff preferences? How to assign the agents according to their skills?
On a second hand, there is a transportation problem. The airport operation being 24/7, the company hires taxicabs for the transportation of it’s staff during night time. To minimize costs, the employees who have the same schedule would share the same cabs. However, this cab sharing has to be done effectively in order not to increase individual travel times. The company used to solve these two problems independently. However, as the Challenge showed, this is far from optimal and huge gains can be made by solving both problems at the same time.
Julio Mejía at el Dorado Airport presenting to the company
50%
decrease in
non-satisfied demand
50%
decrease in number of required vehicles
30%
decrease in transportation travel distances
Winners of the Challenge
Julio Mejía Vera
Daniela Betancourt
Daniela and Julio's software cuts the unsatisfied demand by 50%. Transportation wise their solution reduces by 50% the number of required vehicles and by 30% the overall traveled distances. Their solution manages also to make starting times 40% less variable which is very important for the well-being of the agents. All these results are obtained while maintaining running time below 15 minutes.
David Leonardo Beltrán
Angie Natalia Pineda
David and Angie's software reduces by 30% unsatisfied demand and improves transportation effectiveness by 77%. Part of their heuristic is inspired by a data science algorithm known as K-mean clustering. They also adapted their algorithms to optimize the alignment between demand and agent skills.
María Angélica Piñeros
Daniela Ramírez Alfonso
In just a few seconds their software finds a solution that reduces non-satisfied demand by 30% and requires 60% less cab trips. Their algorithms are designed to narrow the focus on a limited set of fixed schedules so the results are more practical to implement.
Angie Vanessa Barahona
Angie's software focuses on improving the well-being of the agents. It creates a solution that makes space for 11 more rest days. It also reduces the number of required cab trips by 36% with a running time of about 5 minutes. Angie also analyzed the company's solution and proposed organizational changes to reduce unsatisfied demand by 66% while generating 56 more rest days.
A word from the professor
This was a fascinating problem combining transportation and scheduling optimization. Solving these two problems separately inevitably leads to less than optimal solutions. Indeed, optimizing the scheduling problem on its own might result in assigning the same schedules and cabs to agents living far away from each other. Which tends then to increase travel times. In addition, our students had to adequately fine tune their algorithms to take into account many contradictory objectives : maximize service level, minimize transportation costs, minimize individual travel times, minimize the number of changes in schedules and maximize the number of bonus days off!
Prof. Rabie Nait-Abdallah
Leader of the Challenge
Javeriana University