Regularmente los l\u00edderes militares advierten la fragilidad del\u00a0 sistema GPS y durante un tiempo, los investigadores han trabajado en diversas alternativas de soluci\u00f3n, ya sea con\u00a0 relojes cu\u00e1nticos y navegaci\u00f3n inercial.\u00a0La USAF est\u00e1 explorando una nueva alternativa, utilizando los campos magn\u00e9ticos de la Tierra como una forma segura de detectar la ubicaci\u00f3n de las aeronaves y posiblemente otro tipo de\u00a0 veh\u00edculos.\u00a0El Campo magn\u00e9tico terrestre, es una magnitud vectorial. O sea que para cada punto del planeta (en la tierra, aire y mar) tiene componentes vectoriales que son \u00fanicas. Es decir se puede asociar a las tres coordenadas espaciales de un punto las tres componentes del vector Gravedad que pasa por ese punto. El problema es la variabilidad del campo magn\u00e9tico (no es est\u00e1tico en el tiempo: por causas naturales o interferencia del actor). Reemplazar campos magn\u00e9ticos por GPS no es f\u00e1cil. A diferencia de una se\u00f1al clara del espacio, las interferencias electromagn\u00e9ticas del propio avi\u00f3n pueden alterar la capacidad de un sensor para detectar la fuerza del campo. En ese punto, entra la inteligencia artificial, cancelando el ruido de las lecturas del sensor para permitir una mejor se\u00f1al y m\u00e1s precisi\u00f3n.<\/p>\n
Military leaders have been warning of the fragility of GPS for years and researchers have been working on a variety of solutions, from quantum clocks <\/a>to inertial navigation<\/a>. The Air Force is adding a new one, using the earth\u2019s magnetic fields as a secure way to detect location for aircraft and possibly other vehicles.<\/p>\n Magnetic fields emanating from the earth\u2019s surface vary in intensity, just like topography, and so-called magnetic anomaly maps of those fields have existed for years. Back in 2017, Aaron Canciani, an assistant professor of electrical engineering at the Air Force Institute of Technology, set out to see if magnetic sensors (magnetometers) affixed to aircraft could measure the intensity of those magnetic fields and, thus, locate the plane based on where it was in relation to those \u201clandmarks.\u201d His paper<\/a> (and this video<\/a>) shows how to outfit a Cessna plane with magnetometers in the rear and the front. Forty flight-hours worth of data and a lot of work reducing noise from the readings proved the idea viable.<\/p>\n But swapping magnetic fields for GPS isn\u2019t easy. Unlike a crisp clear signal from space, factors such as the electrical operations of the plane itself \u2014 can interfere with a sensor\u2019s ability to detect the strength of the field. This is where artificial intelligence comes in, canceling out the noise from the sensor readings to allow for a better signal and more accuracy.<\/p>\n Researchers in the Air Force\u2019s-MIT Artificial Intelligence Accelerator. community<\/a>, working with scientists at MIT, continued to work on the problem, publishing<\/a> their own paper in July. They showed that magnetic field readings can be accurate to ten meters, only slightly inferior to GPS, which is accurate down to three meters. But magnetometer readings are much less easy to jam than GPS signaling. GPS readings rely on a signal sent along a specific wavelength across vast distances. Magnometers just have to read the magnetic environment around the vehicle.<\/p>\n \u201cBecause of the size of the earth and the magnetic field\u2026 it takes a whole lot to jam a signal coming from the earth, and by a whole lot I mean on the scale of a nuclear blast,\u201d\u00a0 Maj. David “Stitch” Jacobs\u00a0with the accelerator told Defense One.<\/em> \u201cApart from that, it would take a giant scale of a machine to block what\u2019s coming from the earth\u2019s crust. But then you could also cancel it out with machine learning\u201d<\/p>\n The Air Force, working with MIT as part of a new joint accelerator program, <\/a>has posed a challenge to the open AI community<\/a> to help with refined AI tools to improve the magnetic field navigation. The challenge closes on August 28.<\/p>\n In something of an unusual step for the military, Air Force leaders are sharing their dataset with the open research community as part of the program. The government generally doesn\u2019t \u201clike to just give up data,\u201d said Michael Kanaan who directs AI and machine learning for the deputy director of Air Force Intelligence. Instead of using the more burdensome contracts that the military typically uses with defense contractors, Jacobs worked with the Air Force general counsel to create a new sharing license for the program that much more closely resembles the sort that academics use when working on open-source data programs. The clickthrough agreement stipulates that the user has to be using the data for research purposes and will report the results back to the AI community as whole. \u201cThat was something we created just this year and it’s already being executed in several projects, including this one,\u201d\u00a0Kanaan said.<\/p>\n