Mars Express has been orbiting the Red Planet for 12 years. While its controllers know the spacecraft inside out, additional valuable insights may well be hidden within the mounds of telemetry the mission generates – inspiring the first of ESA’s new data mining competitions, open to all.
“The goal of this Mars Express Power Challenge is to predict Mars Express’s thermal power consumption during the martian year ahead, based on its past telemetry,” explains Joerg Mueller, a Young Graduate Trainee in ESA’s Advanced Concepts Team.
“Our intended audience for the competition is the international data mining and machine learning community – whether students, research groups or companies”.
Data mining involves taking large amounts of data and applying sophisticated computer programs to search out previously hidden patterns, associations or anomalies.
The results can be used for machine learning – teaching computers to gain understanding of such patterns – and forecasting the data-generating system’s future performance.
Such an apparently abstract discipline often produces extremely useful results in practical terms, across a wide variety of fields: sifting through astronomical observations for signs of dark matter, smarter text prediction, pinpointing adverse drug combinations within medical data or performing regional flood risk assessment.
“Competitors gain data to hone their techniques on challenging and useful space problems,” comments Dario Izzo, ACT’s scientific coordinator.
“This is the first of a number of planned competitions, in an attempt to demonstrate the usefulness of the approach and to establish a community of participants.
“We’ve established a host website called Kelvins, named after the temperature unit of measurement, with the idea that data mining should aim to reach the lowest possible error in each case – down to absolute zero.”
This first Kelvins competition was developed in close cooperation with the Advanced Mission Concepts and Technologies Office in ESA’s ESOC control centre, which applies new data analysis and computing techniques to help run space missions.
“For the challenge, teams will receive three full martian years of Mars Express telemetry,” adds research fellow Redouane Boumghar. “This will be gigabytes of data – the numerous sensors on the spacecraft can generate new data points every 30 seconds or so.
“Mars Express takes electrical power from its solar array to not only run its platform units and instruments but also to supply actively controlled heaters to maintain an optimal working temperature for the spacecraft.
“As spacecraft systems age, it takes more power to accomplish such tasks, and less power is freely available.
“One of our hopes from the Mars Express Power Challenge is that the mission team can use the resulting predictions to improve power consumption so that Mars Express can end up running longer.”
“Participants don’t need to be space specialists to take part,” adds Dario. “What is termed ‘domain knowledge’ can indeed be useful, but the datasets themselves serve as the starting point.”
The Mars Express Power Challenge takes place between April and August 2016. Check the detailed timeline here.
