Customers from energy to healthcare to communications are leveraging AI to transform their data into gold. If you've used Eureqa and have a story that you'd like to share, don't hesitate to leave a comment!
Originally posted on 10/15/14 by Jess Lin
NASA maximizes flight safety with Eureqa
Commercial airplane travel reigns as the king of safe long-distance travel, but the highly engineered and complex machinery has inherent risks that can be introduced by human error.
In the complex world of an airplane cockpit, Eureqa is helping NASA improve pilots' split second judgement of unexpected flight conditions and reduce flight fatality risk by more than 50%.
Guarantee performance during unexpected conditions
Historically, more than half of all plane crashes have been caused by pilot error; while experience and training reduce risks, even veteran pilots can be fallible under stress. One common cause of pilot error is channelized attention. Similar to both tunnel vision and confirmation bias, pilots faced with sudden and unexpected flight conditions may fixate on certain activities or solutions regardless of the overall situation. Spurred by a 2010 Commercial Aviation Safety Team (CAST) report that analyzed patterns in aircraft accidents over the past decade, NASA has launched a three-year effort dedicated to tackling this deadly problem.
The researchers gathered data from volunteer pilots running high-fidelity flight simulators equipped with about 40 different data recording devices, including eye trackers, electrocardiograms, and motion sensors. But how to turn that data into meaningful action? These sensors record data at a minimum of 60Hz, capturing thousands of data points every 2–3 minutes. Multiplied by diverse pilots, flight tasks, hardware configurations, and flight conditions, the data complexity quickly increases beyond the comprehension of most human minds and analytical tools.
An easy task would be to run a linear regression analysis or analyze each of the sensors individually, but that only gives researchers a black and white view of pilot behavior. The real breakthroughs can only be found by combining all the sensor data to give a full-color, kaleidoscopic view of a pilot's ability to respond to unexpected flight conditions. Identifying the hidden signals within massive volumes of data while juggling multiple pieces of complex, interrelated, and noisy inputs would be a herculean task for most statisticians and tools, so the researchers turned to Eureqa to augment their capabilities.
NASA researchers had used Eureqa in the past for other complex data analysis needs and knew not only could they trust Eureqa to independently search for key patterns and trends important to characterizing pilot engagement, but one of its strongest features allow users to specify computational efficiency. Without the unnecessary complexity typical to machine learning algorithms, calculations can easily be run in real-time, helping researchers quickly test hypotheses against pilot performance across a wide variety of experiments.
Additionally, Eureqa has a unique advantage in providing direct and translatable access to the underlying mathematical expression, instead of just black box algorithms. These transparent results allow users to examine the assumptions behind each of the results and scientifically validate how their calculations would react when faced with new datasets or edge cases. This extra validation step allows users to avoid trusting human lives to unverified theories.
To infinity... and beyond
As the 2010 CAST report discovered, a pilot falling into channelized attention during unexpected in-flight situations can have fatal results. Using AI like Eureqa, NASA researchers can now achieve real-time feedback loops to explore methods that not only safely jog the pilot back into situational awareness, but also prevent channelized attention in the first place. These new methods will be incorporated into next generation pilot training regimens and plane cockpit technologies, leading to permanently safer skies. NASA's research efforts are expected to extend over the next three years as they continue to test and refine the signals that characterize when pilots are engaged, situationally aware, and under high or low workload stress across a wide variety of scenarios.
Challenge: Given datasets with hundreds of thousands of rows and complex, interrelated, and noisy inputs, isolate the signals defining pilot engagement.
Results: Eureqa delivered validated models that could be run in real-time to characterize pilot state data. Quantifying pilot response to unexpected in-flight occurrences enables research into improved training techniques and enhanced cockpit electronics to guarantee peak performance and maximize safety.
Autonomous search across several function sets.
Computationally efficient models that can be run in real-time alongside user behavior.
Scientifically verifiable models which can be cross-checked against new data and edge cases.