AEGIS High-level Scenarios – Scenario #1 – Advanced time-series analytics in the automotive sector

posted in: Blog | 0

In this and the upcoming blog posts, we will present 5 high-level scenarios that try to explain how the AEGIS platform can be used by different stakeholders working in the PSPS domain. The 5 scenarios to be presented, can be found in Deliverable D1.2 and are the following:

  • Scenario 1: Advanced time-series analytics in the automotive sector
  • Scenario 2: Data-enabled services for enriched real-time navigation system
  • Scenario 3: Data-Driven Monitoring and Alert Services for Impaired or High Risk Groups Individuals
  • Scenario 4: Personalised early warning system for asset protection and commercial offering
  • Scenario 5: Open Innovation platform for Data Experimentation and Service Offering

 

Scenario 1: Advanced time-series analytics in the automotive sector

In the next few blogposts we will present concrete examples of how AEGIS will foster data-driven innovation across PSPS sectors, addressing the needs of diverse, in terms of background knowledge and desired output, stakeholders.

The first scenario in this blogpost series describes how AEGIS enables advanced time-series analytics in the automotive sector.

 

The background Alex works in Auto4All, a research centre focusing on environmentally friendly vehicles. His team mainly uses vehicle specifications coming from car manufacturers in order to evaluate their environmental footprint but, until recently, did not have any field data to work with.
The driving need They have now been given access to real time streaming data from 500 taxis, installed by the taxi company that owns them, whose CEO is interested in helping the drivers reduce fuel consumption and improve their driving patterns. For each of the taxis, the installed sensor generates an entry every 10-30 milliseconds (depending on the sensor) with various trip information, including calculated engine load, engine coolant temperature, engine speed, vehicle speed, GPS coordinates, acceleration, gyroscope sensor data etc. The taxis perform transfers mostly in the city of Vienna and each taxi is used by three different drivers, in 8-hour shifts, however there is no entry to indicate when the driver change happens.

Alex is looking for ways to explore and extract insights from the newly acquired data, however he lacks the resources to perform the required time-series analysis and would also like to easily combine them with other data sources, such as weather.

The first steps He has heard of AEGIS and decides to give it a try. He visits AEGIS and sets up a free personal experimentation space where he uploads a small subset of the trip data that correspond to 7 days for 50 taxis. As a start, he wants to create a high-level visualisation of all 50 time-series at once in a map to get a better view of the data. He first performs a time-series reduction to keep one entry every 100 milliseconds and then chooses the option to combine multiple time-series in one and finally selects the map visualisation. He is happy to realise that hard brakes are already evident. Playing around with the map, he filters the results to show one trip per time and selecting the time evolution mode he notices slight changes that could correspond to different driving patterns from distinct drivers.

Confident that the data will reveal useful patterns, he navigates to the AEGIS services and chooses two data-as-a-service offerings to combine with his data:

  1. a weather service providing precipitation, clouds, pressure, temperature and wind measurements, which can be configured based on the desired granularity in terms of location and time intervals
  2. a configurable service providing annotated car accident data from various selected sources and regarding various locations

As both services are free, he imports both as additional input sources for his visualisation and finds the accident indicators on the map in some cases correspond to the locations with hard braking. At the same time, filtering by weather also visually reveals some patterns. In order to further improve the visualisation, Alex chooses to apply the AprioriAll algorithm, which is provided by AEGIS, on the combined (driving and weather) time-series and creates a new visualisation, this time to show only the extracted sequential patterns.

Adoption Phase Excited from the initial results, Alex decides to move from the experimentation level to the full-fledged AEGIS project, where his team will connect all the sensor data streams. The team quickly creates and schedules a braking detection algorithm, an acceleration detection algorithm and a road damage detection algorithm to run every 10 minutes. All they had to do to get them up and running was to apply some rules on top of specific pattern mining algorithms that are by default provided by AEGIS.

After a week of experimentation, Alex and his team finalise the analysis that the taxi company was interested in and proceed to send them a link to the AEGIS interactive maps, created based on the initial Alex’s visualisations, that show under which circumstances the drivers are not driving optimally.

However, Alex and his team have also identified another useful way to leverage the sensor data: after a month of experimenting with the visualisation of the road damage detection algorithm, performing also on-the-field validation of results, the team realises that it provides clear insights into parts of the road network that need repair. Their analysis is so reliable they can even provide alerts when the weather conditions increase the risk of accident above a certain threshold. Although this visualisation no longer holds any driving data, Alex asks the taxi company for consent in order to monetise his team’s finding. Since the company is already an AEGIS service consumer, they happily agree to act as a formal data provider as well for this specific service

Giving back to the community Through AEGIS, Alex and his team publish this visualisation in the form of a weekly and a monthly report and create an alert service where interested parties can subscribe for road risk notifications. These are now available as AEGIS data-as-a-service and visualisation-as-a-service and Alex is confident that the Municipality of Vienna, as well as the road construction companies will be interested in using them.

 

Figure 1: Scenario in a nutshell & key benefits

 

Working on the described scenario has helped us identify some challenges that AEGIS needs to tackle in order to materialize the above workflow. These challenges include the ability to connect streaming data in a way that supports various processing tasks to be applied in (almost) real-time and the need to create terms of usage and licensing schemes for data and services coming from multiple stakeholders.

If you want to learn more about the AEGIS scenarios and the steps we take towards bringing them to life, stay tuned for our upcoming blogposts!

If you can’t wait, you can always start by checking out all our scenarios here.


Blog post authors: NTUA