Mastria: Mobility Orchestration in the digital age

The goal of Mastria was to respond to this problem by monitoring operations in real-time. This was done by integrating the bus network schedule and GPS data, in order to calculate and report delays, positions and incidents. It also included the use of algorithms and smart data analysis (historical and real-time), which made it possible to establish a model of automatic incident detection in the bus network of the Zaragoza metropolitan area. Ten incident categories were defined:

• Bus deviation (e.g. road works)
• Being ahead of schedule
• Being behind schedule
• Late start at the first bus stop
• Loss of connection (bus signal)
• Wrong direction
• Wrong mission
• Trip planned but not performed
• Trip not planned but performed (manual mistake)

Mastria acts as a new layer on the top of each operator´s system to automatically detect the incidents.

In the image above, the route in green is the planned route of the bus, the route in red is the actual route of the bus. Incident Deviation was triggered when the bus was detected outside of the expected path. With Mastria, the authority will immediately call the operator to learn about the reason of the deviation: is there a good reason or has the bus driver decided to change the route? Did the driver mix up the daytime or nighttime route?

The report of categorized incident in real-time made it possible to identify problems and were potentially misunderstood by the operator. The innovation and added value for CTAZ lies in the fact that the platform is accessible by both the authority and the various service operators, which makes it possible to manage jointly in real time all the network information and thereby improve the efficiency of incident resolution , offering an improved service for passengers. Today CTAZ can have real control over these transport operations with real-time visibility.

In Zaragoza, Mastria has shown its advantages in the case of use where the authority must coordinate multiple operators as in Zaragoza. The results of the implementation of Mastria in Zaragoza can be measured in a qualitative way: It is a strong management tool for the city’s mobility supervision authority and gives more clarity and transparency in the cooperation and contract management with the different operators

To solve this challenge, the first step was to understand the problem and define what is exactly happening by modeling passenger flows. To model passenger flows and identify accurately when saturation arrives, Mastria has collected two types of data:

• Ticketing: Check-in and check-out data from ticket gates in order to model origin-destination matrix.
• Signalling data in real-time: Trains position and weight for modelling the people flow in the trains and crossing with ticketing data modelling the entire people flow in the network (including the platforms)

With this input data, Mastria delivered a first level of analysis to better understand the effects of the actions implemented by the operator. The following KPIs were calculated and put in place:

• ¨Fail to Board¨: number and % of people missing a train because it was full (at a platform for each train) – this was also calculated with the origin-destination matrix calculated via information provided by the ticketing.
• Passenger travel time: the main KPI to optimize. The network saturation impacts passengers comfort and travel time, leading some of them to adopt private modes of transport.

It was found that the previous corrective operational actions taken by the operators were not implemented exactly at the right time because the time of saturation of the lines varied greatly from one day to the other (around 30 minutes), without a precise pattern for working days. This variation makes it difficult to schedule the corrective actions in advance. Many factors can influence mobility demand, and this variability makes necessary daily adaptation of the offer.

Once the problem was defined and measured, the second step was to predict the network saturation in real-time.  In order to take the right actions at the right time. Mastria has therefore implemented a prediction of KPIs in real time and a saturation alarm that goes off 30 minutes before saturation thanks to machine learning. An operator in a control room dedicated is then able to take decisions in real-time and can send “empty train” directly to crowded stations without stopping at previous stations. A model built by machine learning algorithms has been implemented in Mastria based on historical data and fed on a real-time basis from signaling sensors.

Results

The use of Mastria has had a real impact on the saturation of Panama metro. First, it has decreased the fail to board during morning peak at interchange station. The number of people who could not board the first available train at the interchange station during morning peak hour has dropped from 17000 to 10000, i.e. from 80% to 41%. It has also improved the waiting time at saturated stations: Pan de Azúcar -295 secs ; San Miguelito -112 secs ; Pueblo Nuevo -146 secs. Thanks to Mastria, average passenger travel time has been reduced by 3 minutes during morning peak hour, improving 30,000 passengers comfort and travel time.

The last step of Mastria which is under development is to give automatic recommendation to the operator to solve the saturation. To do this, the teams of Mastria are working with the operator to build different real-time optimization algorithms.