Bringing Programming and Aviation Knowledge
Together To Create Safer Flight Systems
Computer scientist Carlos Varela has received seed funding from the U.S. Air Force to help make flight data as updated, active, and accurate as possible.
Varela, part of the Data Science Research Center at Rensselaer, will use the more than $100,000 grant to develop sophisticated computer logic programming to help create safer and more efficient flight technology. The grant is part of the Air Force Office of Scientific Research Dynamic Data-Driven Application Systems program.
Flying an airplane takes skill and Varela knows this all too well. The associate professor of computer science is also a licensed pilot. But, what Varela also knows is that flying an airplane also takes data—and lots of it. Pilots are constantly receiving and inputting data from air traffic controllers, weather reports, and multiple sensors throughout the plane to help ensure a safe flight. But what happens when that data isn’t up-to-date, or worse, incorrect?
“Data should drive the flight systems in an aircraft to create active and constantly updated flight data for the pilot,” Varela said. “The idea is that rather than pulling data from sources such as weather forecasting services or air traffic control, which are more static in nature, the system would constantly be comparing and updating source data so that the flight plan is always up-to-date.”
According to Varela, data must understand the connections with other data. For example, we can’t be on time for a meeting if we didn’t get the text message that the meeting was moved to an hour earlier. The data points weren’t connected and we are late for the meeting. In the same way, a pilot or autopilot system cannot take the right action when the data they are receiving is out-of-date or plain wrong. This may have been the problem with the tragic crash of Air France flight 447, which crashed into the Atlantic in June 2009, killing all on board, according to Varela.
The evidence from the crash has suggested that the pilots on board raised the nose of the plane when it shouldn’t have been brought up, he said. A plane with a nose too high will lose speed until eventually it stalls completely. Since the crash, many experts have purported that an equipment failure may have provided inaccurate information to the autopilot, causing it to disengage, said Varela. The pilots may have then incorrectly reacted to the emergency by raising the nose of the plane when in fact it needed to go down to break the stall, according to Varela.
Varela himself experienced this exact type of equipment failure during one of his own flights. Fortunately, Varela was able to determine the failure by comparing bad airspeed data to the ground speed data provided by his own GPS device. It turned out that ice had developed on the airspeed sensor, causing it to fail. Varela did not lower the nose of the plane as he normally would in response to such readings and instead deiced the sensor and safely finished his flight.
These experiences inspired Varela to develop new flight system programming that greatly reduces the possibility of accidents by making connections between the different data streams available to a pilot. This would create redundancies in the data that allow the different data streams to essentially fact-check each other.
An active and redundant flight system may help prevent crashes caused by sensor or other data errors.
The new system will build off what is known in computer science as logic programming by extending a logic programming language to associate probabilities to knowledge. The new system will also give first-class support to redundancy and connections between various spatio-temporal data streams, said Varela.
Varela hopes to create a new system that more easily deals with data streams and quickly admits new data. Such a system could be expanded to include unmanned flight systems and even beyond aviation. The example given by Varela was the citation of scientific findings, where data needs to be well connected to ensure it is accurate and constantly updated as new findings are created.