Inspiration

Who has never suffer any flight’s delay? Currently there are thousands of flying airplanes all around the world. There are many airports, many companies, many different type of planes that makes really hard to find and explain anomalies in real-time. We believe that we can improve flying services for the companies and the users.

What it does

Uses flights information to detect geo-positional, speed and time anomalies and predict them.

This first stage was focused on analyzing flights between LA and NY, assessing the flight path and to detect abnormal movements in flight in relation to its geo-position and heading.

We got two different features.

1) The first allows us to view the history of flights and find anomalous points detected by NUPIC

2) But we also, we want to detect if any current flight is having an anomaly, so we can inform the airline about this situation. To do this, we have created a map having real time data, it’ll notify the user if it find any anomaly.

How I built it

We built our app on top of MoClu (*), that way we can add hardware on demand to deal with tons of data for realtime flights.

On the UI side we are using AngularJS that integrates with Lift 3 and Comet actors, a google maps angular implementation called angular-google-maps and D3 charts.

The backend is built using 2 different web servers.

One of them is used to connect to the models cluster, get and push new flights data coming from external sources also it saves results into MongoDB.

The second web server is used to serve this data to the client. Both web servers can be scaled horizontally.

Challenges I ran into

One problem we noticed is that scalability for this kind of system is crutial.

We detected many anomalies near the airports because we didn't have the enough sample size to learn, this leads to false positives.

Accomplishments that I'm proud of

(*) HTM-MoClu: by definition HTM-MoClu is short for Hierarchical Temporal Memory Models Cluster. Htm-MoClu provides a platform similar to HtmEngine for htm.java applications, and has the ability to scale horizontally using multiple servers.

What I learned

HTM, Sensor configuration, flight data, Clusters configuration.

What's next for Air traffic anomaly detector

A lot of value can be delivered by ATAD in a short-term future, such as:

  • Give the possibility to a person who is about to buy a ticket to select a flight not only based on costs and benefits but also for the possibility of having a fault, making the travel experience better.

  • Find which airports have more anomalies than others, in order to provide this information for later use by airlines or passagers.

  • Report on the number of anomalies by flight to assess their personal onboard.

  • Add weather information to understand it impact and to help plan situations where airports are closed to prevent early collapse secondary airport facilities like hotels.

  • Analyze routes and airports can enable airlines to enhance their routes management in order to reduce costs generated by anomalies.

We believe this information is useful for all the stakeholders, but primarily for the passenger, who will receive a better travel experience.

https://github.com/antidata/ATAD

https://github.com/antidata/htm-moclu

https://github.com/numenta/htm.java

Built With

Share this project:
×

Updates