Agent-based modeling

At the core of the project modeling is agent-based modeling, which considers each passenger as an individual agent with the ability to take decisions that will impact their journey. In turn, other agents’ actions will also have an impact. For instance, heavy road congestion might cause many agents to switch from car to train, which will reduce traffic congestion and increase train crowding. Inside the airport, similar problems can happen at the boarding desks, security checks, etc. In short, this methodology enables us to simulate complex decision-making processes within complex systems.

As IMHOTEP simulates the passengers’ flow under different circumstances during the trip, the focus is on disruptive events and how to manage them. This enables us to compare the base case (‘do nothing’) with various management actions (‘do something’).

Management actions need to be defined before being implemented and ranked. They can be split into many different small actions and combined to produce a diversity of management plans. On top of that, we must also accurately model how the agents (passengers) will react to those actions – more on this later.

The IMHOTEP project envisions three complementary management scenarios:

·      Delayed arriving flights

·      Delayed departing flights

·      A major motorway disruption (such as an incident) near the airport

Management strategies are applied to assess the impact those strategies have in terms of travel time, missed flights, emissions and other KPIs.

Scenario: delayed arrival flight at Palma de Mallorca Airport

To give more detail let us imagine the first scenario of delayed arriving flights, where several flights arrive later than expected during a peak season late afternoon at Palma de Mallorca Airport. This delay might be due to factors such as: air traffic congestion, harsh weather, or accumulation of delays through the day.

Since the airport is likely to be operating near capacity, the late arrivals will stress the facilities and ground handling, as more passengers than expected will crowd the arrival facilities such as passport controls, baggage reclaim, toilets, restaurants, and shops. These passengers will need to reach their destination, either using their own car, or public transport, taxi, shuttle, or a rental car. It might be the case that the buses don’t have sufficient capacity, so how do the agents react? In the ‘do nothing’ scenario, some passengers might choose to wait, but others might take a taxi instead or hire an Uber. This needs to be modeled according to the passengers’ demands.

Another possibility is that there is a long line waiting for taxis, so passengers might then choose either to wait in line or go and get the bus instead. Beyond the ‘do nothing’ scenario, we then need to model the different management scenarios where stakeholders can anticipate and react to the increased demand, e.g., by increasing the frequency of public transportation and hence the capacity or sending more taxis to the airport in advance.

The example described just considers the consequences for arrivals, but some of those arriving late to get their flights, will depart later in the night, resulting in delays for the departures as well, which again is a situation that will need to be managed.

Quantative Management Strategy Selection

Disruption can be described qualitatively, just as we did in the last couple of paragraphs. But how do we select the best management strategy? For that we need to put some numbers (KPIs) based on the agent-based simulation results for the terminal and the airport access and egress. That is exactly what we have done at IMHOTEP, and why we need models for both the access and egress and the airports terminals, to be fed with adequate data.

The Aimsun team has developed a new framework to enable agent-based microscopic solutions. This has been developed as an API in Aimsun Next that reads the passengers demand and manages them through the network until they reach their destination, keeping track of their positions in the network and their travel times.