STA Supporting Material

0015-EX-ST-2019 Text Documents

WaveSense

2019-01-27ELS_221878

Tech Notes www.ll.mit.edu June 2016 Localizing Ground-Penetrating Radar Innovative ground-penetrating radar that maps underground geological features provides autonomous vehicles with real-time localization in all-weather conditions. Autonomous vehicles aim to decrease the number of driver-caused accidents that result in over 30,000 deaths and an estimated 2.3 million injuries in the United States each year. Autonomous ground vehicles (AGVs) could be deployed for military or emergency operations in areas that pose risks to personnel. However, current AGVs are not mature enough for widespread adoption. Most AGV sensors cannot determine the vehicle’s location when adverse conditions, such as heavy rain or fog, snow-covered roads, or lost GPS signals, hamper the functioning of their sensors because those optical sensors rely on traditional roadmap-based information (e.g., lane markers, stop lines). The Localizing Ground-Penetrating Radar (LGPR) uses inherently stable subsurface features MIT Lincoln Laboratory has developed and their geolocation to locate the vehicle even in adverse weather conditions. The prior map a sensor that provides real-time estimates can be seen in gray on the left, the current scan is shown in light blue under the vehicle, and the of a vehicle’s position even in challenging registered data is shown in blue and green behind the vehicle. weather and road conditions. The Localizing Ground-Penetrating Radar a pulse of electromagnetic radiation into a wavelength and that it has sufficient (LGPR) uses very-high-frequency (VHF) the ground and measuring reflections that contrast with the surrounding soil. The radar reflections of underground features to originate from scattering points below the premise of GPR localization is that these generate baseline maps and then matches surface. Reflections occur at the interface subsurface features are sufficiently unique current GPR reflections to the baseline between objects that have different and static to permit their use as identifiers maps to estimate a vehicle’s location. The electromagnetic properties, such as pipes, of the precise location at which their reflec- LGPR uses relatively deep subsurface roots, and rocks in the surrounding “dirt.” tions were collected. features as points of reference because However, it is not these discrete objects but these features are inherently stable and less rather the natural inhomogeneity in subter- Mapping susceptible to erosion or damage over time. ranean geology that often dominates GPR The first step in the LGPR process is to It utilizes VHF radio waves because they reflection profiles. Soil layers and variations develop a map of the environment below the can penetrate rain, fog, dust, and snow. in moisture content cause reflections in the road. In this step, the GPR data of subter- data. Thus, GPR paints a fairly complete ranean “objects” are simply collected along LGPR Methodology picture of the subsurface environment. with GPS tags to form the initial database of For subsurface sensing, GPR is one of the With few exceptions, nearly every discrete subsurface features. This subsurface map is most versatile and prolific sensing modal- object and soil feature is captured, provided then used as a reference dataset to estimate ities today. GPR systems work by sending that it is not significantly smaller than vehicle location on subsequent visits.

Tracking Next, online localization is performed in several steps. When the vehicle is in motion, data are periodically fetched from the database for matching. A 50 m × 50 m × 1–3 m three-dimensional grid of baseline data, centered on a GPS-defined initial location point that is determined by the latitude, longitude, heading, and roll (or tilt) of the sensor, is placed in memory for matching. When the vehicle nears the edge of this grid, it requests a new grid of the same size centered on its new position. In this way, a local grid of baseline data is always maintained. • A search region around the initial The Autonomous Systems Mobile Testbed vehicle has the waterproof localizing ground-pen- location estimate contains “particles” etrating radar array mounted underneath. Additional sensors, such as lidar, camera, and (points on the grid) representing GPS/INS units, are used for verification and studying sensor fusion. candidate locations and orientations. An algorithm iteratively evaluates approximately one-tenth of a center-fre- such a suite is its ability to operate under the particles to narrow the search for quency wavelength. This resolution is conditions that incapacitate other local- the maximum correlation within the finer than typically seen in GPR arrays ization sensors, such as optical or infrared vehicle’s five-dimensional space (easting, and enables high-fidelity matching to systems. Because the LGPR deduces northing, height, roll, and heading). baseline data. location on the basis of stable underground • After several iterations, the highest-cor- • All elements in the array have identical features, it can provide position estimates relation particle is chosen as the most near-field patterns. This requirement even if it encounters severe weather, likely estimate of the vehicle’s current allows path retraversal to resolve obscured or unpaved roads, altered roads, location and orientation. pass-to-pass offset or misalignment. or GPS-denied areas. Fusing LGPR with • The search region is updated and either lidar or other remote sensing methods may expanded or shrunk to reflect this new Demonstrated Capability provide improved localization capabilities estimate. The LGPR system was tested over 100s of for future autonomous vehicles. miles on paved and unpaved roads in four Furthermore, data from below-road LGPR Design U.S. states and was used by the U.S. military features that the LGPR captures can be The basic component of LGPR is a unique to navigate multiple AGVs over more than useful for infrastructure inspection, such as waterproof 12-element antenna array that 1000 miles in Afghanistan’s very demanding finding underground sinkholes or detecting uses a custom VHF stepped-frequency environments. The LGPR method showed structural weaknesses in bridges. Because continuous-wave radar. The VHF system robust performance in these trials. the LGPR compares the data it collects penetrates deeper than typical GPR The article titled “Localizing Ground against prior scans of the terrain, changes systems; thus, it captures deeper, more Penetrating RADAR: A Step Toward Robust in the scans (e.g., subsurface deterioration) stable geological features. Also, because Autonomous Ground Vehicle Localization,” can be readily detected.  VHF frequencies are inherently insensitive published in the January 2016 issue of the to small objects (e.g., a small soda can on Journal of Field Robotics, describes Lincoln Technical Point of Contact the surface will be ignored because of its Laboratory’s demonstration of 4 cm cross- Byron Stanley small VHF radar cross section), their use track localization achieved by a vehicle Control and Autonomous Systems ensures that new reflections are from the driven at 60 mph under fair weather condi- Engineering Group types of geological features cataloged in the tions. Recent work demonstrated real-time stanley@ll.mit.edu baseline data. centimeter-level, highway-speed, nighttime 781-981-4955 The LGPR array of 12 dipoles is linearly localization during a snowstorm that had aligned within a reflective rectangular metal obscured all lane markings. For further information, contact cavity with dimensions of 5 ft × 2 ft × 3 in. Communications and Community Outreach Office Several key modifications to traditional Benefits of LGPR MIT Lincoln Laboratory GPR were fundamental to the design: For an AGV to maintain awareness 244 Wood Street • The cavity depth of 3 inches was of its surroundings and location, it is Lexington, MA 02420-9108 designed for under-vehicle mounting. equipped with a suite of sensors. The 781-981-4204 • The spacing between array elements is main advantage of adding the LGPR to Approved for public release; distribution is unlimited. This material is based upon work supported under Air Force Contract No. FA8702-15-D-0001. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author and do not necessarily reflect the views of the U.S. Air Force. © 2016 Massachusetts Institute of Technology


Document Created: 2016-06-20 13:12:24
Document Modified: 2016-06-20 13:12:24
© 2026 FCC.report
This site is not affiliated with or endorsed by the FCC