Problem

Industry partner: Infty

Infty is a company focused on industrial mathematics, specialising in creating tailored mathematical solutions for industrial partners. Founded in 2018 and based in Sydney, with presence in Poland, the company uses various techniques such as machine learning, artificial intelligence, simulation, statistical analysis, mathematical optimisation, heuristic, and meta-heuristic techniques to discover optimal solutions for industry-related challenges.

Infty works with the focus on the end state on day 1. The commercialisation process begins on the very first day of each project. Key to success is the way Infty brings mathematics and programming to transform businesses to save time and money. Infty has allowed industrial organisations to embed mathematical modelling into their operational infrastructure, covering such diverse topics as forecasting crime propagation, plane tracking and medical research.

Fog and visibility: prediction models

Foreword

Fog affects transportation due to low visibility. The deadliest accident in aviation history, the Tenerife airport disaster, which occurred on March 27, 1977, was caused by a dense fog. Airports procedures’ have improved tremendously over the last 40 years, but the technology to identify fog and estimate visibility used by airports is very localized in space.
A more reliable methods to identify fog and estimate visibility, which connect local and global variables are possible and are sought.

Task

The task is to develop a fog detection and visibility estimation algorithm, which uses:

  • long-term global variables and probabilistic approach for a given location, e.g., temperature, humidity, time of day, etc,
  • local in time variables for a given location, and
  • camera images.

The envisaged solution will couple image-based methods with historical data and data measured at the spot meteorological conditions.

The algorithms and submodels should be developed in open-source languages (R/Python) and well documented (Latex, Markdown). Dashboard implementations, which would demonstrate the solution in action, are welcome, e.g., laptop with camera and artificial fog.

Data

External datasets can be easily generated using commonly available data sources, e.g., Google Earth, Quandl, etc. The company will provide instructions about this to registered participants of the competition.