A Customized AI Planning Solution for Flight Check-in at Brussels Airport: an R&D project
Impact
Preliminary results from this R&D project are encouraging: a drastic reduction in passenger overflow by 94% and queue lengths by up to 79%. This improvement has the potential to enhance the passenger experience and smooth the check-in process in the near future.
Executive Summary
Context
Brussels Airport, a significant air travel hub, embarked on an ambitious R&D project to explore innovative solutions for improving the check-in process during peak times. This initiative aimed to discover new ways to enhance passenger experience and operational efficiency.
Goal of the project
The project's primary goal was exploratory – to determine whether optimizing check-in desk allocations could feasibly reduce passenger congestion and improve the check-in process.
Solution
Radix developed a sophisticated AI solution using a Constraint Programming Satisfiability (CP-SAT) model with over 831,000 variables and 1.73 million constraints. The model utilized flight schedules and estimated passenger arrival patterns to generate optimal desk allocations, ensuring a more efficient check-in process at Brussels Airport.
This case study explores an R&D initiative, not yet in production, focused on potentially transforming the check-in process at Brussels Airport through AI-driven solutions.
Case Study
The challenge
Brussels Airport Company (BAC) faces a significant challenge in managing the check-in process for its passengers, particularly during peak travel times. The airport, equipped with 13 rows of check-in desks totalling 180, has to efficiently allocate these desks to handle approximately 250 flights, serving more than 35,000 departing passengers on busy days.
This complex task presents several key challenges:
Suboptimal Desk Allocations: The check-in desk assignment at BAC is akin to a job shop scheduling problem, a highly complex challenge even for advanced algorithms. The manual approach adopted by the airlines and the BAC check-in team, while functional, sometimes leads to allocations that could be further optimized. This is partly due to the airlines not having a complete overview of the check-in process dynamics, occasionally resulting in queues extending beyond planned areas.
Unexpected Last-Minute Changes: Flight schedules at BAC, typically set in advance, face two main types of last-minute alterations. Firstly, there are changes directly linked to the flight schedule itself, such as delays, grouping, or cancellations, often influenced by external factors. Secondly, changes arise from evolving airline requirements, like adjusting the number of open desks or altering their operating times. These scenarios require the planning team to swiftly adapt, necessitating quick and efficient desk reassignment often at short notice.
Airline Preferences: The assignment of flights to check-in desks can’t be done randomly. Airlines have specific preferences for working at particular rows or desks, further complicating the allocation process.
The briefing
In response to these challenges, Radix was tasked to explore the feasibility of an AI-based solution to optimize the planning of resources for the check-in process at Brussels Airport. The objective was to refine the allocation of check-in desks, focusing on improving resource management efficiency, which in turn would contribute to enhancing the overall passenger experience at the airport.
The work
The solution Radix proposed revolved around a CP-SAT model, a sophisticated mathematical optimization framework. This model was designed to handle the complex environment of check-in desk allocation by considering various variables and constraints.
An illustration of how a flight schedule and knowledge of passenger arrival patterns can be used as inputs for the CP-SAT Solver which produces the desk allocations
Variables: The primary variables were the desk assignments, and 'start' and 'end' times for each flight. The model also considered the number of processed and queuing passengers, as well as passengers outside the queuing zones over time. In total, the model comprised an extensive 831,000 variables.
Constraints: The model implemented various constraints to ensure practical solutions. These included:
Assignment Constraints: Ensuring that assignments didn't span multiple rows and that flights were assigned to neighbouring desks on the same row. Passenger Flow Constraints: Ensuring all passengers were checked in within a specific time frame before departure, and at a particular processing rate. Temporal Constraints: Keeping desks open as required based on passenger flow, without adding or removing desk assignments over time.
Overall, the model incorporated a staggering 1.73 million constraints.
Objective: The ultimate goal of the model was to minimize the passenger overflow – the number of passengers outside of the queuing zones. This was determined by comparing the queue length to the capacity and considering passenger inflow and processing rates.
“The BAC check-in allocation problem really pushed the computational limits and forced us to be creative and design a reliable and scalable solution.”
The future
The successful development of Radix's AI-driven model in the R&D phase has opened up new possibilities for refining the check-in process at Brussels Airport. This innovative approach, though not yet implemented, allows for the simulation of diverse scenarios, offering valuable insights for future enhancements. For instance, it enables the theoretical evaluation of passenger flow impacts resulting from changes like relocating an airline's check-in area or adjusting desk opening times (e.g. 30 minutes earlier). These simulations provide a deeper understanding of potential operational dynamics, highlighting the model's adaptability to different scenarios and its potential effectiveness in an ever-evolving airport environment.
Applicability to other Industries
The potential impact of Radix's AI-based solution for Brussels Airport Company (BAC) in optimizing check-in desk allocations has broader implications beyond the aviation industry. Its applicability extends to various sectors where efficient planning and resource allocation are critical. Here are some key areas and examples where this solution can be instrumental:
Logistics and manufacturing: In industries like logistics and manufacturing, where consumer goods are produced using limited resources and eventually transported over multiple locations, efficient planning is crucial. The same principles used in the BAC case can be applied to optimize production planning, inventory management, routing and distribution networks. By leveraging AI to optimize large amounts of decision variables, companies can significantly enhance their operational efficiency and reduce costs, and consequently improve customer experience.
How CERM defeated today’s printing industry challenges: To tackle the complexity of modern scheduling, CERM contacted Radix to help develop new functionality for their clients' production scheduling module.
Radix, in collaboration with Prof. Dr. Seppe vanden Broucke (UGent), worked together with CERM to develop a bespoke cloud Artificial Intelligence (AI) engine which could be offered to all CERM clients, both improving their production schedule and reducing the manual (human) efforts to schedule. Read our case study here.
Radix and Atlas Copco optimize production planning with AI: the goal of the project was to optimize the production planning of Atlas Copco’s Portable Air manufacturing sites, where complex products are assembled to fit its customers’ requirements. The project aimed to make production planning faster, more efficient and more accurate while reducing human time spent and improving the number of orders delivered on time. Read our case study here.
